Photovoltaic Power Prediction Method Based on Markov Chain and Combined Model

Abstract

Photovoltaic power generation prediction is very helpful for real-time balanced operation in the power market, which in turn will benefit both energy suppliers and customers. Aiming at the prediction of photovoltaic power generation, the paper proposes a prediction method based on Markov chain and combination model. First, this paper preprocesses the data. Then, taking into account the variability and intermittency of photovoltaic power generation, discrete wavelet transform is used to de-noise the data. Next, the paper uses random forest for feature selection to extract feature vectors that are highly related to photovoltaic power generation. Finally, different prediction models are combined, and the weights of different prediction models are dynamically determined through Markov chains. The linear combination of the prediction values of the different prediction models under the dynamic weights is used as the final prediction value. The method in this paper is compared with other methods on real photovoltaic data sets. The experimental results show that the proposed method has a better prediction effect.