Study on scenario generation based on deep clustering and its application in supply-demand imbalance assessment

With the increasing installed capacity of renewable energy generation, how to extract the wind-photovoltaic-hydro-load coupling features effectively and build a joint scenario generation model is the basis for assessing the risk of supply-demand imbalance and operational dispatch. In this paper, deep convolutional embedded cluster analysis theory is introduced to extract potential wind-photovoltaic-hydro-load coupling features from the perspective of data mining, generate joint scenarios, and apply them to the supply-demand imbalance risk assessment of high percentage wind-photovoltaic systems. Firstly, the traditional deep clustering model is improved by using Convolutional Autoencoder with joint loss to solve its problem of insufficient feature extraction and embedded spatial distortion. Secondly, the deep convolutional embedded clustering model is used to cluster wind-photovoltaic-hydro-load time series data and generate wind-photovoltaic-hydro-load coupled scenarios. Finally, a supply-demand imbalance risk assessment system is established to analyze the grid risk evolution path based on the coupled scenarios, so as to provide a theoretical basis for planning decisions for new power systems.