Abstract Aiming at the shortcomings of traditional maximum power point tracking (MPPT) algorithm in photovoltaic system, such as slow tracking speed, easy oscillation and low tracking accuracy, an improved fractional order incremental conductivity (IFOINC) method is proposed. The algorithm combines the variable step-size Peturb and Observe (P & O) method based on power prediction and fractional order conductance increment method. Firstly, the variable step-size P & O method based on power prediction is used to quickly track the MPP, and then the fractional order incremental conductance method is used to further search the MPP. The structure of photovoltaic system, the equivalent model of photovoltaic cells and the DC/DC converter are studied. The IFOINC algorithm is designed, and the MPPT simulation model of photovoltaic system is established. The simulation is carried out in MATLAB/Simulink software environment. The simulation results show that the tracking speed of IFOINC is improved by 42.3% compared with the variable step-size P & O method based on power prediction, and the tracking accuracy of 3.9% higher than that of the variable step-size P & O method, based on power prediction, and the oscillation is eliminated.