The energy-saving effect of the plug-in hybrid electric vehicle (PHEV) is closely related to its energy control strategy as well as the driving cycles. For the plug-in hybrid electric commercial vehicle (PHECV), its driving cycle is relatively fixed. As is the reason, the analysis and prediction of velocity can be fulfilled by historical data and machine learning methods. In this study, the improved velocity prediction method based on Markov chain and back propagation (BP) neural network is initially formulated. The New European Driving Cycle (NEDC) is selected to test the velocity prediction method. The root mean square error (RMSE) of the predicted velocity is 0.0361m/s, 0.1511m/s and 0.4409m/s, respectively, when the prediction time is 1s, 3s and 5s, which illustrates the accuracy and validity of the proposed velocity prediction method. On this basis, the energy optimal control strategy based on velocity prediction is proposed, which is evaluated by comparing with the rule-based energy management strategy and equivalent consumption minimization strategy (ECMS). And the index of price-based equivalent fuel consumption is utilized to evaluate the energy economy. Besides, for a PHECV with P2 configuration, the powertrain simulation model is built with Matlab/Simulink. The simulation is implemented in the driving cycle of NEDC. And the results show that the equivalent fuel consumption using the proposed velocity prediction-based energy optimal control strategy is reduced from 5.3438L/100km and 4.7831L/100km to 4.7334L/100km, with the improvement of energy economy by 9.344% and 1.039%. The proposed velocity prediction-based energy optimal control strategy can provide recommendation for the development of control strategy of PHECV with relatively fixed driving cycles.