Christian Napoli

A radial basis function neural network based approach for the electrical characteristics estimation of a photovoltaic module

The design process of photovoltaic (PV) modules can be greatly enhanced by using advanced and accurate models in order to predict accurately their electrical output behavior. The main aim of this paper is to investigate the application of an advanced neural network based model of a module to improve the accuracy of the predicted output I–V and P–V curves and to keep in account the change of all the parameters at different operating conditions. Radial basis function neural networks (RBFNN) are here utilized to predict the output characteristic of a commercial PV module, by reading only the data of solar irradiation and temperature. A lot of available experimental data were used for the training of the RBFNN, and a backpropagation algorithm was employed. Simulation and experimental validation is reported.

Innovative Second-Generation Wavelets Construction With Recurrent Neural Networks for Solar Radiation Forecasting

Solar radiation prediction is an important challenge for the electrical engineer because it is used to estimate the power developed by commercial photovoltaic modules. This paper deals with the problem of solar radiation prediction based on observed meteorological data. A 2-day forecast is obtained by using novel wavelet recurrent neural networks (WRNNs). In fact, these WRNNS are used to exploit the correlation between solar radiation and timescale-related variations of wind speed, humidity, and temperature. The input to the selected WRNN is provided by timescale-related bands of wavelet coefficients obtained from meteorological time series. The experimental setup available at the University of Catania, Italy, provided this information. The novelty of this approach is that the proposed WRNN performs the prediction in the wavelet domain and, in addition, also performs the inverse wavelet transform, giving the predicted signal as output. The obtained simulation results show a very low root-mean-square error compared to the results of the solar radiation prediction approaches obtained by hybrid neural networks reported in the recent literature.

Simplified firefly algorithm for 2D image key-points search

In order to identify an object, human eyes firstly search the field of view for points or areas which have particular properties. These properties are used to recognise an image or an object. Then this process could be taken as a model to develop computer algorithms for images identification. This paper proposes the idea of applying the simplified firefly algorithm to search for key-areas in 2D images. For a set of input test images the proposed version of firefly algorithm has been examined. Research results are presented and discussed to show the efficiency of this evolutionary computation method.

Using Modularity Metrics to Assist Move Method Refactoring of Large Systems

For large software systems, refactoring activities can be a challenging task, since for keeping component complexity under control the overall architecture as well as many details of each component have to be considered. Product metrics are therefore often used to quantify several parameters related to the modularity of a software system. This paper devises an approach for automatically suggesting refactoring opportunities on large software systems. We show that by assessing metrics for all components, move methods refactoring can be suggested in such a way to improve modularity of several components at once, without hindering any other. However, computing metrics for large software systems, comprising thousands of classes or more, can be a time consuming task when performed on a single CPU. For this, we propose a solution that computes metrics by resorting to GPU, hence greatly shortening computation time. Thanks to our approach precise knowledge on several properties of the system can be continuously gathered while the system evolves, hence assisting developers to quickly assess several solutions for reducing modularity issues.

A Cascade Neural Network Architecture Investigating Surface Plasmon Polaritons Propagation for Thin Metals in OpenMP

Surface plasmon polaritons (SPPs) confined along metal-dielectric interface have attracted a relevant interest in the area of ultracompact photonic circuits, photovoltaic devices and other applications due to their strong field confinement and enhancement. This paper investigates a novel cascade neural network (NN) architecture to find the dependance of metal thickness on the SPP propagation. Additionally, a novel training procedure for the proposed cascade NN has been developed using an OpenMP-based framework, thus greatly reducing training time. The performed experiments confirm the effectiveness of the proposed NN architecture for the problem at hand.

A Hybrid NeuroWavelet Predictor for QoS Control and Stability

For distributed systems to properly react to peaks of requests, their adaptation activities would benefit from the estimation of the amount of requests. This paper proposes a solution to produce a short-term forecast based on data characterising user behaviour of online services. We use wavelet analysis, providing compression and denoising on the observed time series of the amount of past user requests; and a recurrent neural network trained with observed data and designed so as to provide well-timed estimations of future requests. The said ensemble has the ability to predict the amount of future user requests with a root mean squared error below 0.06%. Thanks to prediction, advance resource provision can be performed for the duration of a request peak and for just the right amount of resources, hence avoiding over-provisioning and associated costs. Moreover, reliable provision lets users enjoy a level of availability of services unaffected by load variations.

Some remarks on the application of RNN and PRNN for the charge-discharge simulation of advanced Lithium-ions battery energy storage

In this paper is reported a critical review, experiences and results about state of charge (SOC) and voltage prediction of Lithium-ions batteries obtained by recurrent neural network (RNN) and pipelined recurrent neural network (PRNN) based simulation. These soft computing technologies will be here presented, utilized and implemented to obtain the typical charge characteristics and the charge/discharge simulation procedure of a commercial solid-polymer technology based cell. Simulations are compared with experimental data manufacturers.

Recurrent neural network-based control strategy for battery energy storage in generation systems with intermittent renewable energy sources

The intermittent nature of renewable sources as wind and solar puts a challenge for their use in supply energy to small islands, isolated communities or in developing countries. The integration of battery energy storage system (BESS) or diesel groups is then mandatory. The aim of the paper is to propose a complete recurrent neural networks (RNN) based control strategy of the BESS accounting state of charge (SOC) and terminal voltage and that can be used for their size and to test the use of different type of BESS.

A Cloud-Distributed GPU Architecture for Pattern Identification in Segmented Detectors Big-Data Surveys

Physical surveys in many fields use segmented detectors to sense some physical quantities. In many cases, physical phenomena produce raw data with a rate higher than that possible to identify interesting events and organize data when using a single host. This paper proposes a distributed and parallel processing system that greatly reduces the time needed for identifying events and organizing data. Moreover, our solution can be scaled according to the experiment at hand. This opens the possibility to analyse a wider range of physical phenomena, increases the accuracy of observations, and makes it possible to organize widely distributed experiments involving different laboratories or facilities at the same time. The proposed solution consists of a parallel algorithm able to quickly process patterns sensed by a segmented detector, hence achieving identification and data cataloguing. Moreover, a distributed software architecture properly taps into cloud-based resources to handle the massive amount of raw data generated by an experiment. The results in terms of performances for the proposed distributed and parallel solution have shown a relevant speed up for the required processing.

Improving Files Availability for Bittorrent Using a Diffusion Model

The BitTorrent mechanism effectively spreads file fragments by copying the rarest fragments first. We propose to apply a mathematical model for the diffusion of fragments on a P2P in order to take into account both the effects of peer distances and the changing availability of peers while time goes on. Moreover, we manage to provide a forecast on the availability of a torrent thanks to a neural network that models the behaviour of peers on the P2P system. The combination of the mathematical model and the neural network provides a solution for choosing file fragments that need to be copied first, in order to ensure their continuous availability, counteracting possible disconnections by some peers.