A decision model was constructed to evaluate choices between nine diff
erent nutrient recipes for use in a controlled environment greenhouse.
Probabilities of occurrence for different solar irradiance ranges wer
e developed from historical data for use in the model. The Beta functi
on is used for computing the probability density function of the diffe
rent solar irradiance ranges. Irradiance levels are separated into cat
egories according to the weather forecast data available on public tel
evision each day. An important part of the model is its ability to lea
rn from the data collected each day. Once a forecast is made, its vali
dity can be tested at the end of the day, when the actual solar irradi
ance values are known. At the end of each day, the new Beta parameters
are derived using the new sample mean and variance. The Beta paramete
rs are updated in this manner for each weather scenario, adding one mo
re data set each day. These values were incorporated into the model so
that the relevant probability density function can be modified each d
ay to reflect more accurately the conditions prevailing at each partic
ular site. The model is being implemented to drive a computer-controll
ed multihead nutrient injector for use in growing hydroponic crops.