Power Allocation for Islanded Microgrid Under VSG control with BI-LSTM Neural Network

This paper deals with the power allocation problem for an islanded microgrid based on the virtual synchronous generator (VSG) control with the predication provided by the BI-directional long short-term memory (BI-LSTM) neural network. A VSG control strategy is employed to improve system stability by simulating the external and droop characteristics of synchronous generators. Different nameplate capacity of each VSG has a serious impact on the reactive side power distribution. For addressing this problem, a virtual parallel capacitor is first employed to reduce power allocation errors on the reactive side. Next, a prediction mechanism by means of BI-LSTM neural network is proposed to determine the upper and lower limits of power. Finally, a simulation example with detailed discussions is provided to demonstrate the effectiveness and applicability of the scheme developed in this paper.