A real-time control-oriented mean value engine plant model that includes engine thermals and cold starts is developed for a Toyota Prius 2015 plug-in hybrid engine in Modelica and MapleSim and validated experimentally. The model consists of an engine block model, intake and exhaust manifold models, and a throttle model. An advantage of the engine block model is the ability to compute the frictional Mean Effective Pressure during engine cold starts from calculated air, oil, and coolant temperatures at various locations in the engine block. Traditionally, engine thermals are modelled utilizing thermal resistances and capacitors. The proposed model utilizes linear graph theory with terminal equations to study the topology of the different components that affect engine thermals, including engine head, liner, coolant, and oil sump. Linear graph theory is introduced as a methodological tool able to represent the various components included in the thermal engine model, reducing the complexity of automated differential-algebraic equations generation. The generated model equations are solved using a generic solver. The throttle model is extended to include reverse air flows, extending the optimization range in model predictive controllers. Bench tests are conducted on a Toyota Prius engine where the flow rates, temperatures, and pressures are measured over the engine air path. The respective temperatures and pressures are measured over the different engine components along with the engine torque and speed for different engine settings. Experimental values are utilized to estimate various parameters for the new engine models. The developed model is integrated with an engine manifold model that includes 1-D spatial variation developed by integrating Orthogonal Collocation with the Method of Characteristics. The manifold model solves the one-dimensional Euler equations used to model compressible quasi-one-dimensional flow with heat transfer and friction effects. Discharge rates and engine friction mean effective pressures are compared with experimental data. In summary, a validated real-time engine model that captures important dynamic phenomena and suitable for control applications is developed, allowing simulations with other engine air path models created using Modelica.