It is difficult to accurately estimate the state of charge (SOC) of a vanadium redox flow battery in a DC microgrid energy storage system. Thus an SOC estimation method based on hybrid coyote optimization with grey wolf optimization algorithms (HCOAG) to optimize a kernel extreme learning machine (KELM) model is proposed. First, the improved COA (ICOA) and simplified GWO (SGWO) algorithm are fused by a sinusoidal crossing strategy to form an HCOAG algorithm. This is used to optimize the parameters of the KELM model. Then, a benchmark function is used to test the HCOAG algorithm, and the optimization ability of the HCOAG algorithm is compared with other intelligent algorithms. Finally, the accuracy and feasibility of the estimation method are verified by simulation and experiment on a CEC-VRB-5 kW battery. The results show that the estimation accuracy of the proposed HCOAG-KELM method is better than that of the GWO-KELM, ICOA-KELM, KELM, unscented Kalman filter (UKF) and extended Kalman filter (EKF) algorithm models, and the SOC estimation error is within 2%, which can meet actual demand.