Massive Data-Driven Joint Characterization Model for Multi-Dimensional Cluster Characteristics of Energy Storage Power Stations

New energy storage cluster consists of a large number of energy storages with various types. The operational parameters of different clusters are significantly differentiated, which makes it difficult to analyze the multi-dimensional characteristics directly and quantitatively. Therefore, characteristics of energy storage clusters are unavailable for the dispatching side, which affects the accurate and effective regulation and control of energy storage resources to a certain extent. In order to make multiple types of energy storage power stations participate in power system dispatching better, it is necessary to mine their characteristics from massive data. This paper proposes a joint characterization model for multi-dimensional characteristics of energy storage power stations. Firstly, this method detects and corrects anomalies in heterogeneous measurement data from multiple sources based on the random forest technique. Secondly, the massive data are divided into measurement priorities according to data types and periods, and mapped to sampling frequencies. Moreover, the models of indices are accurately estimated through stratified sampling and plurality regression. Finally, a multi-dimensional model of energy storage power station characterized by multi-dimensional triplets is constructed and implemented on real energy storage power stations.