A configuration has been devised which allows for the testing of estimation and control algorithms in an experimental system which realistically simulates an exothermic reaction taking place in a batch or continuous stirred reactor. In the PARSEX (Partially Simulated Exothermic) reactor, flows, levels and temperatures are measured experimentally and live steam is injected at a rate calculated to simulate the exothermicity of the reaction. In this work, temperature control of a batch reactor has been studied using Generic Model Control(GMC). Since, in most industrial applications, the state variables required for controller implementation are not all measurable or, not with sufficient accuracy for control purposes, a state estimation technique has been applied as well. In this work, the amount of heat released by the reactions has been estimated online using an extended Kalman filter, and incorporated into the GMC algorithm. Simulation results had shown that the Kalman filter gave an accurate estimate of the amount of heat released and together with the GMC controller, gave reliable robust control. An experimental extension of the work using the PARSEX reactor shows that the extended Kalman filter is rather more sensitive to plant/model mismatch than would have been predicted from simulations alone; these directly affect the performance of the GMC controller.