Application of time-series methods to disturbance estimation in predictive control problems

Disturbance effect is a very important issue to be analyzed and considered in any control problem. Models representing the disturbance effect on the process variable are usually obtained for this purpose. These models are useful to design, for instance, classical feedforward compensators. However, when predictive control strategies are being used, no only the model capturing the disturbance effect is necessary, but that future estimations of the disturbances would be also extremely useful to get advance in the prediction capabilities. Measurable disturbances can be represented as time-series data and processed in order to estimate future variations. Therefore, this paper is focused on studying how well-known time-series methods can be used to obtain disturbance forecasts in order to be taken into account in predictive control algorithms. The methods used here have two main objectives: to detect the nature and properties of any variable represented as time-series data, and second, to forecast future behavior based on acquired data series. An example using the solar radiation measured from an industrial greenhouse is presented to show how these future estimations can be obtained.