Considering the regular characteristics of city bus routines, dynamic programming (DP) is much better than rule based strategy in developing the global optimal control strategy for a dual-motor coaxial coupling propulsion electric bus (DMCEB). However, its complexity and computational burden hinder the implementation capability. To solve the trade-off problem between optimum and practicability, a novel systematic extraction method of energy management strategy is proposed in this paper. Globally, DP is applied offline to find optimal operating points of propulsion components. Explicit shift schedule lines are designed by utilizing nonlinear support vector machine (SVM) with a new efficient factor tuning method. Torque split ratio (TSR) is extracted statistically in quadratic functions by using recursive k-means clustering and piecewise polynomial fitting methods. Then integrated rule-based strategy can be developed systematically and implemented online with proper operating range and limits. Eventually, hardware-in-the-loop (HIL) experiment results demonstrate that the proposed method can be executed through a 16-bit onboard processor in real-time. The online performance of the improved strategy is highly consistent with simulation results, which can reduce energy consumption by 21.06% compared to the preliminary heuristic strategy. Moreover, the applicability of the proposed methodology is further verified over different typical driving cycles.