The use of direct drives in linear and rotary axes as well as increased power density of main drives offer the potential to raise feet rate, acceleration and thus allow higher productivity of machine tools. The induced heat flow rates of these drives could lead to thermo-elastic deformations of precision related machine tool components. In order to reduce thermally caused displacements of the tool-center-point and to prevent a negative impact on the achievable accuracy, the induced heat flow rates of main drives must be dissipated by effective cooling systems. These systems account for a major share of the machine tool’s total energy consumption.With the intention to overcome the area of conflict regarding productivity and energy efficiency, a so called thermoelectric self cooling system has been developed. To convert a proportion of thermal losses into electrical energy, thermoelectric generators are placed in the heat flow between the primary part of a linear direct drive and the cooling system. The harvested energy is directly supplied to a pump of the water cooling circuit, which operates a decentralised cooling system with reasonable coolant flow rates. For predicting the thermoelectric system behaviour and to enable a model-based design of thermoelectric self cooling systems, a thermal resistance network as a system simulation in MATLAB/Simulink is presented. The model is applied to a feed unit with a linear direct drive and allows the calculation of harvested energy as well as the simulation of steady and transient states of the cooling system. The comparison of simulative and experimental determined data indicates a predominantly high model prediction accuracy with short simulation times. At an early stage of development the model turns out to be a powerful tool for design and analysis of water flow thermoelectric self cooling systems.