Abstract: This paper presents a composed piecewise function digital predistortion model based on vector switch model to address the problem of substantial nonlinearity and strong memory effect in wideband power amplifier (PA). This paper sets two thresholds according to the PA characteristics and decompose the input signal into sub-signals by using two thresholds. In low stage, we use a generalized memory polynomial (GMP). A memory polynomial (MP) model was used in the middle stage and a higher-order GMP model was used in the high stage. By sharing the fundamental MP and configuring various cross terms, memory depth, and nonlinear degree for each segment, the proposed model improves linearization performance while also reducing the complexity of parameter extraction and the impact of the piecewise function′s discontinuity close to the piecewise point. The model also offers greater flexibility in dynamic threshold adjustment for various PA manufacturers. In simulation, the proposed model can reduce adjust channel power ratio from -34.1 dBc to -46.5 dBc in comparison to MP model improving 3.5 dB. Error vector magnitude performance is improved from 6.46% to 1.06%, or roughly 0.7% better than MP model. Compared to the GMP model, the complexity of parameter extraction has been reduced by 77%.