In this work the inverse model based temperature control of a batch chemical reactor is presented. The controller combines feed-forward and feedback elements. The feed-forward part is the constrained direct inverse, while multiple feedback structures are studied. Since we have a model with good correspondence with the physical system, the inverse of it can produce the manipulated variable. The feedback is needed to eliminate the effect of modeling errors that is especially important in steady-state. For the compensation of modeling error the following channels can be used: the manipulated variable, the set-point and the model parameters. The difference between each channel is how direct they act, and how fast can the consequences of manipulation be seen. These channels do not exclude each other, usually a faster feedback and a slower adaptation is applied where the control problem requires it. The example on which the controller strategies are presented is a 1 liter batch chemical reactor. The reactor has duplicated wall, and a laboratory thermostat supplies cooling/heating medium that flows through on the jacket of the reactor. The manipulated variable is the set-point of the thermostat, and the controlled variable is the inner temperature of the reactor. Simulation experiments support that the developed controller strategy is superior to conventional feedback loops, as it contains handling of measured disturbances, manipulator constraints, dead time and nonlinearity.