1. Antonio Romero Jiménez. Sistemas y análisis de la Información geográfica. 2º.RA- MA.2008. 2. David W. Allen. Getting to know ArcGis ModelBuilder.1ESRI. 2009. 3. ALEXOPOULOU, S.: “A decision support system for the integration of renewable energy sources into water desalination systems (REDES)”, en Renewable Energy Development European Conference and APAS-RENA Contractors Meeting. Florencia: EDIFIR.1996. 4. Amador Guerra, Julio. Análisis de los parámetros técnicos en la aplicación de los sistemas de información geográfica a la integración regional de las energías renovables en la producción descentralizada de electricidad . 2000. 5. P. J. Ribeiro Jrand P. J. Diggle, “geor: A package for geostatistical analysis,” R news, vol. 1, no. 2, pp. 14–18, 2001. 6. P. Diggle and P. J. Ribeiro, Model-based geostatistics. Springer, 2007. ABSTRACT Designing an optimal network measurement sensors for monitoring geographical wind resource: Statistical techniques and GIS Laura Frías Paredes ([email protected]) / Edurne Pascual Chahuan / Martín Gastón Romeo National Renewable Energy Centre (CENER) Fermín Mallor (Public University of Navarre, Spain) PO.010 EVALUATION OF THE STATISTIC INDEX OF FEASABILITY A DEEPER DESCRIPTIVE ANALYSIS BY MEAN OF BICLUSTERING FUTURE DEVELOPMENTS : BICRITERIA INDEX AND PARETO CURVE EWEA Resource Assessment 2015 – Helsinki – 2-3 June 2015 This work presents a new methodology developed to design a sensor network extent to ensure a certain level of uncertainty in the development of spatial analysis of long-term estimations of wind energy resource. The resource maps are a useful tool in the screening process areas of interest for the use, development and installation of renewable energy. Different techniques and sources of information are used to estimate the long-term resource availability in a geographic area, weather models, satellite images and ground measurements are the most common. Obviously, the more accurate information is provided by real measurements using appropriate sensors, but this information presents the disadvantage of referring to a particular point and the usual scarcity of this type of data in many parts of the world. On the other hand, it should be noted that in the development of resource maps the actual available information is introduced to adjust estimates and validate the results. In this work it is considered an initial estimation resource in a geographic area, with the goal of identifying the ideal location for a sensor network that allows us to reduce the uncertainty of the estimate. For this purpose, the spatial and temporal variability is analyzed by using spatial statistic and to identify distinct areas and representative points. This analysis is merged with the rest of geographic information available by mean of a GIS to obtain the most interesting locations to the measurement network installation. The methodology developed in this work can play an important role when the potential of the wind energy is analyzed in large geographic areas. REFERENCEs For future developments we could perform more complex and exhaustive area optimization developments taking into account not only linear approach distances but path costs calculation for displacement for measuring towers and site installation. Other factors can be added such as view shed analysis, airport entrance areas, etc. Along this work, we will use a database of wind resource over Iberian peninsula obtained with the numerical weather prediction model SKIRON with a resolution of 0.05º. Period: 2003-2013 (hourly wind data). The first task is to analyze the data from the geostatistical point of view. Biclustering consists of simultaneous partitioning of the set of samples and the set of their attributes (features) into subsets (classes). Samples and features classified together are supposed to have a high relevance to each other. A typical situation to calculate bicluster are a high dimensional dataset with many variables, so that normal cluster algorithms lead to diffuse results due to many uncorrelated variables. Also biclustering is useful if there is a assumed connection of objects and some of the variables in the dataset, e.g. some objects have 'similar' patterns for a given set of variables. EVALUATION OF THE ECONOMIC FEASIBILITY INDEX FOR EACH POINT GIS is a set of tools to evaluate geographical data, characterized by being geo-referenced, with the ultimate purpose of making territorial studies, obtaining analysis using spatial relationships between elements themselves in the territory. Today its use is more and more important because of all the features that are made in this tool: operations between layers of information from different sources, storage, data unification and visualization of different variables of meso- microscale models that characterize the resource (vm, power density) of a given territory .The objective is to obtain optimized areas for wind farm construction and installation of measuring towers. To do this we consider various aspects with different legal, economic and environmental criteria, but always the determining factor will be the existing resource. These factors have been taken into account in Spain: 1. Wind resources obtained by SKIRON. 2. Minimum legal distance and maximum economical distance to national and autonomic roads. 3. Maximum economical distance to high-voltage power lines. 4. Maximum economical distance to substations. 5. Absolute restriction of protected areas. 6. Minimum legal distance to population centers. 7. Maximum economical slope obtained from the digital elevation model. So, each grid point has associated a score that represents its economic feasibility. This score has been obtained by mean of a ponderated sum of the different layers and taking into account that a score equals to 0 means that the site is not suitable by some reason. In this case the objects to classify are the grid points meanwhile the variables are the monthly mean of wind speed. We can seen 3 different areas, two of them present high relationship in months from November to March (2 and 3). However the bicluster 1 presents high relation when summer months are taken into account. The design of an optimal network is related with minimizing the variance of our estimator. Now, the data is prepared to analyze the variance using a variogram. The variogram is the fundamental tool to describe the spatial covariance. The figure shows the adjust of model to the empirical variogram. For our case, r 2 = 0.3481, σ 2 = 0.2370 and the range is h r = 1.57. These parameters give the amount of variance that it is explained when a sensor is located at a certain point. Near the point, the variance decrease and the influence is lost at distances greater than h r . The variogram could be considered as a reference of the amount of variance that should be explained where a sensor is located at a certain point. In the following strategies are considered to select the network grid. Basically, all of them selects a point, then downscale the variance surface according with the spatial dependence given by the variogram and so on. In this analysis the whole points of the area are scored according to the order that they are selected by the methodology. At this point each pixel of the area is identified by two different scores. The goal is to minimize both index to identify the most interesting sites. The pareto curve appears as an interesting tool to achieve this goal The key idea of this method is to use the variance surface of the remote monitoring data to find those sites with higher variance which will be the candidates to be included in the network. Once a location is included, the variance is downscaling in the vicinity according with the estimated spatial dependence detected in the remote monitoring data.