The power allocation of composite energy storage in microgrid has always been a hot research topic in the industry. Taking photovoltaic microgrid as the application scenario, the nonlinear characteristics of power fluctuation and load fluctuation in time will affect the control of energy storage system. To solve this problem, the composite energy storage control strategy based on source-load power prediction was optimized, and the fuzzy algorithm was used to deal with nonlinear problems. The fuzzy controller was added into the power distribution of energy storage elements by taking into account the nonlinear relationship between energy storage battery SOC and the power fluctuation of the microgrid. The difference distribution algorithm between the predicted energy and the actual grid-connected energy of the microgrid was designed, and the reference value of active power of the energy storage system was corrected in real time, so as to achieve the effect of regulating the SOC of the energy storage system. In the long-time scale, the actual grid-connected power can accurately track the grid-connected scheduling, and the energy storage elements will not be overcharged or over discharged, thus reducing the damage rate of the energy storage elements, improving the safety and stability of the composite energy storage system, and prolonging the service life of the energy storage system. The experimental results show that compared with the control strategy before optimization, the fluctuation range of battery SOC can be reduced by 15.6% under the same working conditions, and the fluctuation range remains between 40% and 60%.