Estimating reference crop evapotranspiration using HGA-LSSVM

Reference crop evapotranspiration (ETo) is the basis for estimating crop evapotranspiration and for computing crop irrigation requirements. In recent years, Least squares support vector machines (LSSVM) have been applied to forecasting in many areas of engineering. In this paper, a novel hyper-parameter selection for LSSVM regression is presented based on hybrid genetic algorithm (HGA). The HGA not only has the advantage of global searching of GA, but also the advantage of local optimization ability of Levenberg-Marquardt optimization algorithm. The LSSVM is applied to the forecasting of reference crop evapotranspiration (ETo). Three ETo prediction models of different meteorological factor input were established based on HGA-LSSVM. These models were verified by measured meteorological data. The ETo computational results by three models were in accordance with the measured results. It also indicated that three ETo prediction models based on LSSVM had the strong predictive ability. And three models predictive ability was 5 factor input LSSVM-ETo-1> 4 factor input LSSVM-ETo-2>3 factor LSSVM-ETo-3 in turn. So HGA-based hyper-parameter selection for LSSVM regression and LSSVM applied to ETo forecast are feasible.