Black start path optimization is a key step in the process of power grid restoration. To solve the problems of large search space, high computational complexity, and slow convergence speed in black start path optimization of new energy power grids, an improved A * algorithm based black start path optimization strategy for new energy power grids is proposed. Firstly, the uncertainty of the output power of new energy stations is analyzed, and a BP neural network is used for ultra short term power prediction. Secondly, taking the restoration time of transmission lines as the objective function, a black start restoration optimization model is constructed considering constraints such as line operation status and starting power. Based on this, the traditional A * algorithm is improved by introducing weight coefficients and path smoothing strategies to obtain the optimal restoration path for black start of power grids containing new energy. Finally, a MAT/AB simulation example is used to compare the performance of different path optimization strategies during the black start process. The results show that this algorithm has higher recovery efficiency compared to other strategies, and improves the available active power and total power generation during the black start process, effectively ensuring the reliability and speed of the power supply scheme.