Abstract: The capacity configuration of electrochemical energy storage power stations requires comprehensive power system data. This paper proposes a novel adaptive configuration method for electrochemical energy storage power station capacity based on parallel hybrid attention. The method calculates the appropriate capacity of energy storage power stations using historical data, analyzes the relationship between the operational characteristics of energy storage power stations and grid demands, sets constraints for capacity configuration, and employs a parallel hybrid attention mechanism to integrate capacity configuration features. This approach identifies a cost-optimal capacity configuration scheme that meets system requirements, enabling adaptive configuration of electrochemical energy storage power station capacity. Experimental results demonstrate that the proposed method exhibits excellent performance in terms of the key evaluation indicator of peak-valley difference in photovoltaic power output,approaching the actual maximum peak-valley difference of the photovoltaic output power. It achieves “peak shaving and valley filling” in power output, reduces the operational pressure on generating units, and enhances the overall operational stability of the energy storage system.

    Key words: parallel hybrid attention; electrochemical energy storage power station; capacity adaptive configuration; recurrent neural network(RNN); convolutional neural network (CNN)