Abstract:
A greenhouse is a controllable dynamic system, covering a range of characteristic times.
Two approaches to greenhouse-environment optimization have been tried: (1) Mimicking the actions of expert growers, and (2) model-based optimization.
The former approach is the traditional one, which cannot readily adjust to new circumstances.
Model-based optimization requires reliable models, as well as good computational methods.
The maximum principle method produces costate variables which can be used as meaningful control-policy parameters.
Sometimes a brute-force search in control space is used.
Whole season control trajectories show how emphasis sometimes shifts from one variable to another.
It turns out that the costate trajectories are not strongly dependent of weather.
Hence, costate sequences of past years can be used as control policy for future years.
Prices may have a significant effect on the costate.
Greenhouse dynamics need only be considered when weather fluctuations have a characteristic period commensurate with the time constant of the greenhouse.
Obtaining optimal solutions under these circumstances, requires a detailed weather forecast, which is usually not available.
|