This doctoral thesis systematically considers the problem of deadlock occurrence and prevention in the discrete event systems. The deadlock is a state of the system which does not allow any action. The subject of research has been the method of synthesising deadlock prevention supervisory controller in automated systems which are modelled by Generalized Petri nets and Labelled Petri nets. The thesis presents and compares the method of synthesising deadlock prevention supervisory controller by automata and Petri net theories. The analysis shows Petri net advantages i.e. a smaller number of discrete states in the discrete event model and an easier method of calculating deadlock prevention supervisory controller. The thesis proposes the method of simplification from the complex Petri net of discrete event system which consists of controllable and uncontrollable transitions to the simple Petri net which consists solely of controllable transitions, and which is suitable for synthesising deadlock prevention supervisory controller. The simplification of complex process Petri net is done by the bottom-up synthesis method in combination with methods of reduction and the decomposition of complex Petri net. Further, conditions for achieving liveness and a deadlock in the Petri net are being defined. The thesis proposes an algorithm for the design of deadlock prevention supervisory controller. This algorithm combines both the method of iterative minimal siphon control which searches for minimal siphons using the matrix method of searching for all siphons in the net and gradually from the set of all siphons extracts those which are not minimal, and the method of analyzing reachability graph. The result of the deadlock prevention algorithm is the set of constraints on the reachable and initial marking of Petri net by means of linear inequalities. Constrains on the reachable markings determine the number of control places which are included in the state deadlock prevention supervisory controller, and which are connected to the process Petri net. The proposed algorithm enables the calculation of a more simple deadlock prevention supervisory controller that has a fewer control places than the algorithms that have been proposed so far. The verification of algorithms is shown on two processes. The first is a flexible batch process in the pharmaceutical industry and the second is a maritime traffic system. Flexible batch process which is considered in the thesis consists of several different shared resources belonging to the same type in which, in case of resource breakdown, the jobs are allocated to the other resource of the same type. The deadlock situation can occur in case of resource breakdown if several different products are being produced at the same time. By application of deadlock prevention supervisory controller, which can be obtained by the method proposed in the thesis, it is possible to continue production of several different products at the same time without the deadlock, even in the situation of the resource breakdown. Regarding the other process, maritime traffic system, the deadlock situations can occur when channel and pool systems which accommodate vessels sailing in opposite directions are improperly used. The thesis illustrates the formal method for design of the deadlock supervisory controller which controls traffic signalisation in the channel and pool systems in a way to prevent a possible deadlock, and with the least possible disturbance of the on-going traffic.