ABSTRACT: The interacting multiple model algorithm was a hybrid estimation method using multiple model tracking the state of nonlinear systems. Through the measurement, it is verified that the photovoltaic power generation array had measurement noise. In the partially shadow cases, it had many characteristics. They are nonlinear, time-varying uncertainty and more local power peak. The improved interacting multiple model algorithm was adopted to improve the accuracy of tracking power curve. The multiple models switching probability was used to identify the multiple local power peak to avoid misjudgment problem. The new technique proposed was able to quickly and accurately track the maximum power point, regardless of whether the shadow cases. The efficiency of photovoltaic power generation was improved. The correctness and validity of the new control strategy were verified by the simulation and the experiment.