Methodologies for estimating missing sensor data in order to assure proper functionality of a monitoring and control system are presented. Short term forecasting models for learning the dependencies between the measured sensor values are analyzed. Experiments were performed in Matlab for selecting a model and its associated structure appropriate for deployment on an embedded module. Once the structure is fixed, the embedded module is then responsible for estimating the model parameters and forecasting the environmental variables for a short time horizon.
Short term estimation of environmental variables for improving the fault tolerance of distributed control networks. Nicoleta Stroia, Daniel Moga, Alexandru Lodin