Reconfigurable intelligent surface (RIS) is seen as a promising technology for next-generation wireless communications, and channel modeling is the key to RIS research. However, traditional model frameworks only support Tx-Rx channel modeling. In this letter, a RIS cascade channel modeling method based on a geometry-based stochastic model (GBSM) is proposed, which follows a 3GPP standardized modeling framework. The main improvements come from two aspects. One is to consider the non-ideal phase modulation of the RIS element, so as to accurately include its phase modulation characteristic. The other is the Tx-RIS-Rx cascade channel generation method based on the RIS radiation pattern. Thus, the conventional Tx-Rx channel model is easily expanded to RIS propagation environments. The differences between the proposed cascade channel model and the channel model with ideal phase modulation are investigated. The simulation results show that the proposed model can better reflect the dependence of RIS on angle and polarization.