Application of Least Square Support Vector Machine in Photovoltaic Power Forecasting

To reduce the impact of random power output of photovoltaic(PV) system on power system,a PV out put forecasting model based on energy storage and least square support vector machine is built.This model can forecast PV power out one hour ahead and optimize the installed capacity of energy storage device that compensates the difference between expected value of PV output and its actual output value.A forecasting model of solar irradiation on ground surface that reflects the variation information of cloud layer is presented;then adopting generated energy of PV system,solar irradiation on ground surface and air temperature series,two least square support machine models are built by unified modeling and time series modeling respectively;the trained two models are tested and assessed under different day-types to verify the effectiveness of the proposed algorithm and built models.Results show that the built models not only can well solve the randomization problem of PV system,but also can reduce the installed capacity of energy storage devices.