With technological advances, the industrial sector is using enabling technologies based on the concepts of Industry 4.0, among them the modeling of systems based on Digital Twins, which allow building models of real physical systems from field measurements, with the ability to follow the dynamic system changes. This article describes the development of a digital twin for a cooling system based on data from the sensors of a thermal power plant. The digital twin created used data from a plant currently in operation at the major system, and the automatic rules extraction approach from the test database was used to form the model's knowledge database. The principal component analysis technique was used to reduce the dimensionality of the system to reduce the computational effort of the model. An algorithm for automatic update of rules during the operation of the digital twin was also proposed, making the model learn during the operation even without the need to retrain and, consequently, reduce error of the model's response in the short term.