In modern petrochemical plants the corporate and societal demands to plant safety and minimum environmental effects are high. These demands rise high performance requirements to the technical systems, specially the process control and safety systems including an effective operator support system with fault detection capability. The systems must have high reliability also against erroneous operations which may cause shutdown situations or quality deviations. PVC producing plants have high inherent hazard potentials. The risk can be minimised to an acceptable level with an integrated and well structured system consisting of a highly reliable basic process control system, a supervisory control system, an operator support system and a high integrity protective system. The PVC reaction studied in this paper is exothermic and non-linear, and open-loop unstable. This may cause stability problems with varying operational conditions. The size of modern PVC reactors is so big that two independent cooling systems are needed to keep the reaction under control. The parallel control of two such cooling systems is complicated. Many existing solutions have problems with oscillatory control and even stability problems, causing reduced safety margins. In this paper a system to solve the stability problem of parallel control is suggested. The philosophy of the solution is decoupling of the dynamic dependency of the parallel control loops. The system is as shown simulated on real process data, and it is installed on a Norsk Hydro PVC plant. When a reaction comes out of control, or rather when it shows a tendency to do so, an operator support system is needed to give the process operator an early warning so he can overrule the automatic control before a shut down situation occurs. Here is described a system for monitoring the heat of reaction in the autoclave. The system consists of a model based estimator. The process model is based on first principles. There are uncertainties in some parameters, specially the heat transfer and the measurement of the cooler difference temperature. The operator support system performs state and parameter estimation as well as fault detection, based on an extended Kalman filter. The system is tested on-line. In this paper results of simulations and on line estimates are presented.