A voltage control method for isolated grid based on long short-term memory neural network optimized by improved particle swarm optimization algorithm

With the access of numerous distributed generations such as wind turbines, photovoltaic generators, and energy storage devices, the composition and operation mode of the power grid are becoming increasingly complex, and more uncertain factors affect the recovery of the isolated grid voltage. Aiming at the problem of optimal control of isolated grid voltage under uncertain factors, a voltage control method for isolated grid is proposed, based on long short-term memory neural network optimized by improved particle swarm optimization algorithm. Firstly, the relationship between the gain of the distributed power controller and the performance of the isolated grid controller is analyzed. Then, the particle swarm optimization algorithm is utilized to optimize the hyper-parameters such as the learning rate and neuron number of the long short-term memory neural network, so as to prevent the hyper-parameters from selecting training effect affecting network improperly. Finally, the optimized long short-term memory neural network is mapped to the relationship between the voltage deviation of the distributed generation and the controller gain, and a voltage optimization control model for the isolated grid is constructed. The simulation results show that the voltage of the distributed generation is stabilized at 1 after 0.92 s, which indicates that the proposed optimal control method improves the voltage level of the isolated grid, verifying the effectiveness of the method.