J. Aguilera

Generation of hourly irradiation synthetic series using the neural network multilayer perceptron

In this work, a methodology based on the neural network model called multilayer perceptron (MLP) to solve a typical problem in solar energy is presented. This methodology consists of the generation of synthetic series of hourly solar irradiation. The model presented is based on the capacity of the MLP for finding relations between variables for which interrelation is unknown explicitly. The information available can be included progressively at the series generator at different stages. A comparative study with other solar irradiation synthetic generation methods has been done in order to demonstrate the validity of the one proposed.

Recurrent Neural Supervised Models for Generating Solar Radiation Synthetic Series

In this paper, a neural network method for generating solar radiation synthetic series is proposed and evaluated. In solar energy application fields such as photovoltaic systems and solar heating systems, the need of long sequences of solar irradiation data is fundamental. Nevertheless those series are not frequently available: in many locations the records are incomplete or difficult to manage, whereas in other places there are no records at all. Hence, many authors have proposed different methods to generate synthetic series of irradiation trying to preserve some statistical properties of the recorded ones. The neural procedure shown here represents a simple alternative way to address this problem. A comparative study of the neural-based synthetic series and series generated by other methods has been carried out with the objective of demonstrating the universality and generalisation capabilities of this new approach. The results show the good performance of this irradiation series generation method.

RMS Current of a Photovoltaic Generator in Grid-Connected PV Systems: Definition and Application

This paper includes a definition of a new and original concept in the photovoltaic field, RMS current of a photovoltaic generator for grid-connected systems. The RMS current is very useful for calculating energy losses in cables used in a PV generator. As well, a current factor has been defined in order to simplify RMS current calculation. This factor provides an immediate (quick and easy) calculation method for the RMS current that does not depend on the case particular conditions (orientation, location, etc.). RMS current and current factor values have been calculated for different locations and modules.

Characterization of Concentrating Photovoltaic modules by cooperative competitive Radial Basis Function Networks

Concentrating Photovoltaic (CPV) technology attempts to optimize the efficiency of solar energy production systems. As conventional Photovoltaic (PV) technology, suffers from variability in its production and needs models for determining the exact module performance. There are several problems when analyzing CPV systems performance with traditional techniques due to absence of standardization. In this sense it is remarkable the importance for the emerging CPV technology, of the existence of models which allow the prediction of modules performance from initial atmospheric conditions. In this paper, a CPV module is studied by means of atmospheric conditions obtained using an automatic test and measuring system developed by the authors. The characterization of the CPV module is carried out considering incident normal irradiance, ambient temperature, spectral irradiance distribution and wind speed. CO^2RBFN, a cooperative-competitive algorithm for the design of radial basis neural networks, is adapted and applied to these data obtaining a model with a good level of accuracy on test data, improving the results obtained by other methods considered in the experimental comparison. These results are promising and the obtained model could be used to work out the maximum power at the CPV reporting conditions and to analyze the performance of the module under any conditions and at any moment.

MEFES: An evolutionary proposal for the detection of exceptions in subgroup discovery. An application to Concentrating Photovoltaic Technology

Subgroup discovery is a broadly applicable data mining technique whose main objective is the search for descriptions of subgroups of data that are statistically unusual with respect to a property of interest. The obtaining of general rules describing as many instances as possible is preferred in subgroup discovery, but this can lead to less accurate descriptions that incorrectly describe some instances. Under certain conditions, these incorrectly-described instances can be grouped into exceptions. A new post-processing methodology for the detection of exceptions associated to previously discovered subgroups is presented in this paper. The purpose is to obtain a new description to improve the accuracy of the initial subgroup and to describe new small spaces in data with unusual behaviour within the subgroup. This post-processing methodology can be applied to the results of any subgroup discovery algorithm. A post-processing multiobjective evolutionary fuzzy system is developed following this methodology, the Multiobjective Evolutionary Fuzzy system for the detection of Exceptions in Subgroups (MEFES). A wide experimental study has been performed, supported by statistical tests, comparing the results obtained by representative subgroup discovery algorithms with those obtained after applying the post-processing algorithm. Finally, MEFES is applied in a real problem related to the description of the behaviour of a type of solar cell in the Concentrating Photovoltaic area providing useful information to the experts.

Niching genetic feature selection algorithms applied to the design of fuzzy rule-based classification systems

In the design of fuzzy rule-based classification systems (FRBCSs) a feature selection process which determines the most relevant features is a crucial component in the majority of the classification problems. This simplification process increases the efficiency of the design process, improves the interpretability of the FRBCS obtained and its generalization capacity. Most of the feature selection algorithms provide a set of variables which are adequate for the induction process according to different quality measures. Nevertheless it can be useful for the induction process to determine not only a set of variables but also different set of variables. These sets of variables can be used for the design of a set of FRBCSs which can be combined in a multiclassifler system, improving the prediction capacity increasing its description capacity. In this work, different proposals of niching genetic algorithms for the feature selection process are analyzed. The different sets of features provided by them are used in a multiclassifier system designed by means of a genetic proposal. The experimentation shows the adaptation of this type of genetic algorithms to the FRBCS design.

Feature Selection Algorithms Applied to Parkinson's Disease

In Parkinsons Disease an analysis of Medical Data could highlight some symptoms, which can be used as a complementary tool in an early diagnosis. This paper analyses some Filter and Wrapper Feature Selection Algorithms and combinations of them that determine some relevant features in relation to this problem. The experimentation carried out with a data set of patients allows us to determine a set of different premorbid personality traits that can be considered in the early diagnosis of Parkinsonism.

A neural network approach for generating solar irradiation artificial series

In this paper a relevant problem in the photovoltaic solar energy field is considered: the generation of artificial series of hourly solar irradiation. The proposed methodology artificially generates series following the average tendency of the hourly radiation series k t in a given place. This is obtained by making use of a set of historical values of this series in such place (for training purposes) as well as the daily clarity index K T of the year to be generated. This information is employed for the supervised training of a proposed neural network model. The neural model employs a well known paradigm, called Multilayer Perceptron (MLP), in a feedback architecture. The generation method is based on the MLP ability to extract, from a sufficiently general training set, the existing relationships between variables whose interdependence is unknown a priori. This way, the presented design methodology can implicitly include all the available information. Simulation results show the good performance of the irradiation series generator, and the general applicability of this methodology in the estimation of highly complex temporal series.

A method for evaluating shading losses involved in a GC-BIPV using real data

This paper presents a simplified method for evaluating the energy loss involved in a grid connected building integrated photovoltaic (GC-BIPV) system due to partial shading. The method consists in the comparison of the monthly theoretical curve (for un-shaded PV generator) and real fitted curve (deducted from monitoring data) of the maximum energy produced by the PV generator.

An evolutionary fuzzy system for the detection of exceptions in subgroup discovery

Subgroup Discovery (SD) is a data mining technique whose main objective is the search for descriptions of subgroups of data that are statistically unusual with respect to a property of interest. General rules describing as many instances as possible are preferred in SD, but this can lead to less accurate descriptions that incorrectly describe some instances. These negative examples can be grouped into exceptions. The paper presents a new evolutionary fuzzy system for the detection of exceptions associated to rules previously obtained by a SD algorithm. Considering the initial subgroup and associated exceptions, the aim is to obtain a new description in order to increase the accuracy of the initial subgroup. This algorithm can be applied to the results of any SD algorithm. An experimental study shows the utility of the proposal, which is also applied in a real problem related to concentrating photovoltaic technology, providing useful information to the experts.