Control of the underexcitation mode of synchronous machines in power plants

The problem of underexcitation in synchronous machines (alternators) becomes particularly relevant in controlling electric power plants to prevent instability and superheating of the machines. In addition, in extreme but actual conditions, the machine loses the synchronism with respect to the electric network. Effective conventional solutions (which involve classical PID controllers) do not always solve the problem since the controller action may be not sufficiently prompt to properly face the disturbances. In this paper we propose an hybrid controller consisting of an integral action with gain tuned by a recurrent neural network. With a suitable design of the network structure the final controller´s performances are better than those of the classical one