Under the premise of satisfying steady operation of distribution network, fully utilizing the rated capacity of the grid-connected inverter, and excavating the reactive power support of photovoltaic power station, which can promote economic operation of distribution network. Thus, it is of great practical significance to research the evaluation method of reactive power support for photovoltaic power station. Firstly, the evaluation method based on dynamic time warping(DTW) algorithm and the evaluation index system is proposed,so as to realize reactive power support evaluation in the real time. Then the prediction method of photovoltaic output based on deep belief network(DBN) and the predictive index system is proposed, combined with predicted data to realize reactive power support evaluation in the future state. Numerical examples show that the predicted evaluation result is highly consistent with the actual evaluation result,which verifies the effectiveness of the established evaluation and prediction method.