Sverdlova A. Information and measurement system for diagnosing complex thermal power facilities using retrospective information

The dissertation is devoted to the improvement of the process of diagnosing and forecasting the technical condition of service stations through the development of models, methods and information-measuring system for diagnosing the condition of complex thermal power facilities using current and retrospective information. The functional scheme of the diagnostic process and the general mathematical model of the dynamics of changes in time of the measured quantity values in the form of a vector random field at fixed spatial coordinates are proposed in the work. A method of diagnosing elements of complex thermal power facilities is proposed, which is based on current and retrospective information, which allowed to compare the forecast indicators with existing ones, as well as to take into account previous experience of operation of complex thermal power facilities. The structure, the scheme, the experimental sample of the module of multilevel system are made and experimentally checked. The structure of the neural network is based on a combination of algorithms of recurrent neural network with long short-term memory and autoencoder, which allowed to predict equipment failures in a small number of anomalous precedents and increased the reliability of predicting abnormal states by 9%.