The demand for information transmission is ever increasing. Low-voltage power line communication has become one of the preferred solutions to solve the "last kilometer" problem. However, this communication mode is limited by the lack of bandwidth resources and there is no efficient bandwidth allocation strategy. In order to solve this problem, a dynamic bandwidth allocation for Power Line Communication (DBA-PLC) based on an optimal stationary model is proposed. DBA-PLC takes communication quality evaluation as the decision variable, adopts a parallel training method of an optimal stationary model to allocate bandwidth adaptively, and optimizes the algorithm by data filtering. Through simulation experiments, the feasibility of the strategy is verified. It is concluded that when the data filtering rate is 20%~30%, an ideal bandwidth allocation result can be obtained under with less resource occupation. It is proved that the bandwidth allocation result has continuous effectiveness when the channel change rate is less than 0.25, that is, updating the bandwidth allocation strategy once in a certain period of time can ensure the high bandwidth utilization of communication. This work is supported by the Natural Science Foundation of Shandong Province “Study on Wavelet Basis-based OFDM Grouping Strategy in the Low-voltage Power Line High-speed Digital Signal Transmission” (No. ZR2010FM041).