Aiming at the problem of the prediction accuracy for day ahead short- term radiationintensity of photovoltaic power plants, and considering the learning parameters of support vectormachine(SVM) have great influence on the performance of predicting model, in order to improvefurther the accuracy of prediction data, introduces the cuckoo search algorithm to optimize the penaltyfactor c and the kernel parameter g of SVM, puts forward a combined forecasting method based oncuckoo search and support vector regression. The simulation result shows that this method canimprove greatly the prediction accuracy and precision of photovoltaic radiation intensity, it ' s feasibleand efficient, can apply to photovoltaic online prediction.