This paper presents a detailed model of a DC train to be applied to an unified AC/DC power flow analysis. The model will consider the regenerative braking of the train and the squeezing control that derives part or all regenerated power to the rheostatic braking system depending on the catenary voltage. The model also considers an on board accumulator (ACR) that can be charged with the regenerated power, depending on the network and train parameters. When the train is in traction mode, both the network and the ACR can contribute in feeding the train traction system. 2 In the present work, the authors develop a train model to be combined with the previous mentioned power flow approach. The proposed model combines the regenerative braking with an on-board accumulator, the so-called (ACR) developed by CAF Company. ACR is a Spanish acronym that stands for to Acumulador de Carga Rapida (Fast Charge Acummulator). The train can work under two different modes. The first one is the traction mode. In this mode, the required power can be provided by the catenary or the ACR depending on the ACR charge level, the amount of power demanded by the train and the catenary voltage. The second one is the braking mode, in which the power can be injected in the catenary, used to charge the ACR or burned in the rheostatic braking system depending on the operating variables. The squeezing control is also simulated when the catenary voltage exceeds a given value. Thus, part of the power is derived to the rheostatic system, and over a given catenary voltage value, no power can be injected in the catenary and all regenerated power must be burned. the developed model is a general parametric model that could be used with most of the accumulation technologies.The authors must remark that this model was developed for power flow purposes, that is why the model do not include any derivative term considering the train dynamics or the electrical network dynamics. A more accurate formulation considering such dynamic behaviour will make the formulation much more difficult and it wont add more information or accuracy when calculating the power flow solution. The authors run a steady state simulations at each simulation instant neglecting the transients between successive instants as it was proposed in [8]. This approach, known as stationary equivalent method for moving loads, is widely accepted among the authors, not only for modelling DC traction networks, but also for modelling AC traction networks for high speed trains [2][3][4][5][9][10][11][12][13][14]. This method assumes that the speed of the trains is not very high to induce pronounced electrical transients and the dc traction network slowly moves from one state to another as the locations and the input power of the trains vary. For this reason, steady state problems are solved at each instant, neglecting the electrical transients and the dynamic of the components. It is true that this is a simplification of the reality, and a differential model solved for transient purposes would be more acc...