This thesis deals with the optimal temperature control of an injection molding machine. Based on a nonlinear mathematical model with four in- and outputs and 38 states, different control strategies are designed. A static nonlinear feedforward control is developed and combined with a Multiple-Input-Multiple-Output Proportional-Integral (MIMO-PI) and a Linear-Quadratic-Regulator (LQR) as feedback controller. For the LQR, an integral part is added to avoid a steady control error if there are model parameter uncertainties. To implement an LQR, all of the states of the model have to be known. Since only four states out of 38 are measurable, a Kalman filter is designed. To determine how the controller reacts in the presence of variations of several model parameters, a sensitivity analysis is presented. After that, the designed controller and observer are tested with several meaningful test cases and compared with a state-of-the-art control strategy. This analysis is the starting point for the estimation of those parameters, which show the highest influence on model behaviour. For this task, the Kalman filter is extended by a parameter estimator and an optimization problem is formulated to identify unknown parameters. The investigated controller and state observer are finally tested with a model with parameter uncertainties.