A review of measure-correlate-predict (MCP) methods used to estimate long-term wind characteristics at a target site

Abstract

So-called Measure-Correlate-Predict (MCP) methods have been extensively proposed in renewable energy related literature to estimate the wind resources that represent the long-term conditions at a target site where a short-term wind data measurement campaign has been held. The main differences between the various MCP methods lie fundamentally in the type of relationship established between the wind data (speed and direction) recorded at the target site and the wind data recorded simultaneously at one or various nearby weather stations which serve as reference stations and for which long-term data series are also available. This paper reviews a wide range of MCP methods that have been proposed in the context of wind energy analysis, a number of which have been implemented in wind energy industry software applications. This review includes the initial methods first proposed in the 1940s which generally attempted only to estimate the long-term mean annual wind speed from a single reference station, and extends up to the most recent methods proposed in the present century based on automatic learning techniques which use several reference stations. In addition to offering a description of the linear, non-linear and probabilistic transfer functions used by the different algorithms, the hypotheses on which these functions are based and the data format with which the various methods work (time series or frequency distributions), this review will also cover limitations in the use of MCP methods, the uncertainty associated with them and the different reference data sources that have been studied. In this sense, the extensive collection of MCP methods which is brought together and reviewed in this paper, ranging from the simplest and easiest-to-use models to the most complicated computational ones which require specific user experience, comprises an extremely useful catalogue when it comes to choosing the best predictor method.