The invention provides a photovoltaic ultra-short-term combined prediction method based on an Adaboost algorithm, and belongs to the field of photovoltaic power calculation, and the method comprises the steps: obtaining the prediction accuracy of each sub prediction model for each sample point, and calculating the weight of each sub prediction model in a combined model according to an error rate;Updating the weight distribution of the training samples; According to the updated sample point weight, the error rate of each sub-prediction model is recalculated, iteration is ended when a preset iteration termination condition is met, and a final combined prediction model including the sub-prediction models is obtained; And according to the finally obtained combined prediction model. Differentweak learners are trained for the same training set, and then the weak learners are integrated to form a stronger combined prediction model. By introducing a prediction standard reaching rate concept,taking prediction standard reaching as a target and combining an Adaboost algorithm idea, combined optimization of a plurality of weak learning algorithms is realized, prediction risks of a single prediction method can be avoided, prediction precision is improved, and stability of a prediction model is enhanced.