Aiming at the problem of low prediction accuracy of single power prediction model due to the impact of photovoltaic power fluctuation

a combined photovoltaic power prediction model based on similar day clustering is proposed. Firstly

k-means clustering is selected to divide the original power data into three similar day sample sets of sunny

rainy and cloudy according to different weather types

and the variational mode decomposition(VMD) is used to decompose the similar day samples; Secondly

the convolution neural network is used to optimize the support vector machine(CNN-SVM) and bidirectional short-term and short-term memory(BiLSTM) neural network

respectively

to predict and superimpose the decomposed power data and combine the prediction results with weights

and the grid search algorithm(GS) is used to find the optimal combination weight to improve the performance of the combination prediction model. Finally

the validity of the PV power prediction model proposed in this paper is verified by the one-year measured data of a photovoltaic power station in Australia. The experimental results show that the model proposed in this paper can predict the photovoltaic power well and has strong adaptability no matter what weather type.