Since the wireless sensor network has appeared, it plays an important role in human life. It has the ability to help people automatically sense the real world, where is inaccessible before. Although the applications of wireless sensor networks are increasing nowadays, one of the great issues facing sensor network is the lifetime of a system. Since the system relies on the limited battery, finite lifetime is its biggest challenge. Fortunately, solar energy harvesting for the outdoor environment is promised to address that problem. Compared with that, indoor energy harvesting is still an immature field for most applications, since it is complex to estimate the available energy for indoor lights. To complement the vacancy of indoor energy harvesting field, improve the performance of the traditional estimation model, a new solar cell model is proposed to substitute the traditional solar cell model. The traditional estimation model is used to roughly dimension the indoor system, regardless of the different impact generated by different light sources. The new estimation model, was established under the real indoor environmental conditions. A classification model is created to distinguish the sort of the light sources. The single solar cell of the traditional model is replaced by 5 different solar cell models fitted for each kind of light source. With that, the system has the ability to select the most suitable solar cell model based on the classification result. Moreover, a verification model was built to evaluate both estimation models. The evaluation result shows that the new model has the ability to perform well under changing light condition.