The control process in compound control systems uses system state predictions in short, intermediate and longtime periods. These predictions are used in order to design corresponding controllers. Data evolution in such systems can be presented in time series format that can be used for state prediction. These time series represent the results of interactions between subsystems of the compound system. The time series can be formulated as a result of a three-step process. At the first step the signals are digitized and decomposed by means of spectral analysis and digital filtration. At the second step the neural network that is used for signal prediction is built. At the third step machine learning algorithms and the multi-component neural network are used to obtain the state/signal prediction in a variety of time scales. This analysis is used to design controllers.
Paper
Full text
State Prediction in Compound Control Systems via Time Series: Neural Network Approach
Semantic Scholar · Engineering · 2019
Abstract
The control process in compound control systems uses system state predictions in short, intermediate and longtime periods. These predictions are used in order to design corresponding controllers. Data evolution in such systems can be presented in time series format that can be used for state prediction. These time series represent the results of interactions between subsystems of the compound system. The time series can be formulated as a result of a three-step process. At the first step the signals are digitized and decomposed by means of spectral analysis and digital filtration. At the second step the neural network that is used for signal prediction is built. At the third step machine learning algorithms and the multi-component neural network are used to obtain the state/signal prediction in a variety of time scales. This analysis is used to design controllers.