Search academic texts

Information × Registration Number 0221U107099, 0119U101859 , R & D reports Title Development of a system for monitoring the level of harmful emissions of TPP and diagnosing the equipment of power plants using renewable energy sources on the basis of Smart Grid with their collaboration popup.stage_title Head Babak Vitalii P., Доктор технічних наук Registration Date 26-12-2021 Organization Institute of Engineering Thermophysics of NAS of Ukraine popup.description2  The object of research - a system for monitoring the level of harmful emissions of thermal power plants and diagnosing the equipment of power plants using renewable energy sources, based on Smart Grid The purpose of the work is to develop the diagnostic system and its software for the equipment of power plants that use renewable energy sources. The analysis of methods of intelligent data analysis for forecasting has been carried out, the methods of predicting anomalous states of complex technical objects with the use of deep learning algorithms with LSTM and auto-coder architectures have been further developed, which makes it possible to develop an efficient, data-driven system that can process events that happened thousands of discrete time steps ago and remember them. The practical value of the results is that for the first time, an information-measuring diagnostic system was developed using neural network technology, which allows you to evaluate the current state of work, detect anomalous states in time, prevent possible forced shutdowns, predict and plan maintenance measures. A sample of the measuring module, which can include sensors of those physical quantities that are used when diagnosing a specific system, is developed. The software has been developed that allows unifying data formats and data exchange protocols for different types of sensors and means of communication of subsystems, as well as between subsystems. An experimental study of neural network models was carried out. A variation of training parameters for neural network models with different learning periods, number of hidden layers, and neurons are formed. A program of neural networks training has been developed. Checking the results of an experimental study of the equipment failure prediction system in real conditions showed an increase in the reliability of prediction of anomalous states by 9%. Product Description popup.authors Yeremenko Volodymyr Stanislavovych Zaporozhets Artur O. Sverdlova Anastasiia Dmitrivna popup.nrat_date 2022-03-09 Close