56 research outputs found
A spiking neural network-based long-term prediction system for biogas production
No full textEfficient energy production from biomass is a central rarch issue in the context of clean alternative energy resource. In this work we propose a novel model based on spiking neural networks cubes in order to model the chemical processes that goes on in a digestor for the production of usable biogas. For the implementation of the predictive structure, we have used the NeuCube computational framework. The goals of the proposed model were: develop a tool for real applications (low-cost and efficient), generalise the data when the system presents high sensitivity to small differences on the initial conditions, take in account the “multi-scale” temporal dynamics of the chemical processes occurring in the digestor, since the variations present in the early stages of the processes are very quick, whereas in the later stages are slower. By using the first ten days of observation the implemented system has been proven able to predict the evolution of the chemical process up to the 100th day obtaining an high degree of accuracy with respect to the experimental data measured in laboratory. This is due to the fact that the spiking neural networks have shown to be able to modeling complex information processes and then it has been shown that spiking neurons are able to handle patterns of activity that spans different time scales. Thanks to such properties, our system is able to capture the multi-scale trend of the time series associated to the early-stage evolutions, as well as their interaction, which are crucial in the point of view of the information content to obtain a good long-term prediction
Small Lung Nodules Detection based on Fuzzy-Logic and Probabilistic Neural Network with Bio-inspired Reinforcement Learning
No full textInternal organs, like lungs, are very often examined by the use of screening methods. For this purpose we present an evaluation model based on a composition of fuzzy system combined with a neural network. The input image is evaluated by means of custom rules which use type-1 fuzzy membership functions. The results are forwarded to a neural network for final evaluation. Our model was validated by using X ray images with lung nodules. The results shows the high performances of our approach with sensitivity and specificity reaching almost 95% and 90% respectively, with an accuracy of 92.56%. The new methodology lower considerably the computational demands and increases detection performances
A neuro wavelet-based approach for short-term load forecasting in integrated generation systems
No full textAdvanced and Adaptive Dispatch for Smart Grids by means of Predictive Models
No full textAbstract—Integrated generation systems are increasingly con- sidered suitable to supply remote areas, less developed coun- tries, and small isolated communities with power. The energy management investigated in this paper concerns a smart grid encompassing a photovoltaic park. We propose a novel cloud- distributed solution to determine the best energy dispatch, i.e. where energy is going to be used and whether to change the operating points for some consumption devices. Neural networks have been used to predict both energy production and consump- tion, making it possible to strategically set the activation time of loading devices and to minimize energy flow changes. Moreover, cloud computing resources make it possible to have fast and distributed computation on the big amount of data gauging power production and consumption
An advanced neural network based solution to enforce dispatch continuity in smart grids
No full textIn energy generation systems including a photovoltaic park, fluctuations are the norm: both production and demand levels can vary on hourly basis. Hence, energy management and dispatching systems have to cope with the possibility of inadequate production while satisfying as much as possible user demands. We put forward a management solution that models the behaviour of each production plant and consumption device, and determines energy allocation. For this, gathered data are wavelet transformed to let us retain only the useful characteristics of data on both large and small scales of the signal. Models are handled by several neural networks which perform predictions in advance of 48-hour, with a granularity of half an hour. Moreover, according to realtime user demands, the management solution determines energy flows between production plants and consumption devices. Therefore, while in some cases it might be necessary to postpone the activation of some consumption devices, in others we can take advantage of a production surplus. Thanks to the proposed solution proper actuators can be programmed beforehand to improve the fairness to users, and use peaks of energy production, thus reducing green energy shortage, and extra costs
Organic solar cells defects detection by means of an elliptical basis neural network and a new feature extraction technique
No full textThe study proposed in this paper devises to develop a new methodology based on elliptical basisneural network (EBNN) and on a new feature extraction technique in order to recognize theorganic solar cells (OSCs) defects. The feature extraction procedure has been obtained by usingthe co-occurrence matrices and the SVD decomposition applied to atomic microscope forceimagery. The polymer-based OSCs used for this work have been produced at the optoelectronicorganic semiconductor devices laboratory at Ben Gurion University of the Negev. The testsperformed show that with our approach it is possible to obtain a correct classification percentageof 95.4% proving that the proposed feature extraction technique based on the co-occurrenceMatrix and the SVD decomposition is very effective in the detection of different types of OSC surface defects
A new design methodology to predict wind farm energy production by means of a spiking neural network-based system
No full textIn this paper, a spiking neural network–based architecture for the prediction of wind farm energy production is proposed. The model is also able to evaluate the wake effects due to interactions between the elements of a wind farm on the energy production of the whole farm. This method has been applied to a large wind power plant, composed of 28 turbines and 3 anemometric towers, located in the rural area of Vizzini's
unicipality in province of Catania, Italy, that is characterised by a complex orography and an extension of 30 km2. For the implementation of this architecture it was used the “NeuCube” simulator. The results show that the presented method can be successfully applied for predictions of wind energy generation in real wind farm also in presence of faults.Depto. de Estructura de la Materia, Física Térmica y ElectrónicaFac. de Ciencias FísicasTRUEpu
A New Iterative FIR Filter Design Approach using a Gaussian Approximation
No full textThe paper presents a novel iterative methodology for the design of FIR filters based on an approssimation of the desired filter frequency response using a Gabor system generated by the Gaussian function. The proposed method exhibits simplicity of implementation, comparable to that of window-based design methods and ensures accuracy in the fulfillment of design requirements, comparable to the one achieved by the Parks- McClellans method. Furthermore, two other advantages of this method are: closed-form formula for the tap coefficients of the filter and the smooth, monotonically decreasing behaviour of the frequency response from DC to infinite frequency
A novel cloud-distributed toolbox for optimal energy dispatch management from renewables in IGSs by using WRNN predictors and GPU parallel solutions
No full textIntegrated generation systems (IGSs) are todayincreasingly considered to exploit renewable energy in order tosupply load for remote areas, less developed countries and smallisolated communities. The IGS investigated in this paper enclosea PV park with battery energy storage system and this configurationis considered as case study for the campus called Cittadellaat the University of Catania. A novel cloud distributed toolboxfor the purpose of an optimal energy dispatch management byusing Wavelet Recurrent Neural Networks (WRNN) predictorsand Graphic Processing Units (GPU) parallel solutions in theIGS considered as case study is proposed. Results coming fromthe implementation of the proposed cloud architecture are herepresented
Automatic heart activity diagnosis based on Gram polynomials and probabilistic neural networks
No full textThe paper proposes a new approach to heart activity diagnosis based on Gram polynomials and probabilistic neural networks (PNN). Heart disease recognition is based on the analysis of phonocardiogram (PCG) digital sequences. The PNN provides a powerful tool for proper classification of the input data set. The novelty of the proposed approach lies in a powerful feature extraction based on Gram polynomials and the Fourier transform. The proposed system presents good performance obtaining overall sensitivity of 93%, specificity of 91% and accuracy of 94%, using a public database of over 3000 heart beat sound recordings, classified as normal and abnormal heart sounds. Thus, it can be concluded that Gram polynomials and PNN prove to be a very efficient technique using the PCG signal for characterizing heart diseases
- …
