Articles
BELL PEPPER PRODUCTION PREDICTION BASED ON COLOUR DEVELOPMENT DISTRIBUTION, SOLAR RADIATION AND GLASS HOUSE TEMPERATURE DATA
Article number
674_46
Pages
375 – 380
Language
English
Abstract
To optimise logistics, establish better prices and make better deals with retailers, growers of bell peppers and their packing houses have a need for a prediction of the production of, say, the next week or two.
A method was developed to predict bell pepper harvest with a time horizon of one week and two weeks on a greenhouse level.
In a first stage several fruit colour development models were proposed.
A first group consisted of ARX models taking into account the greenhouse temperature and solar radiation data.
Because of some model inadequacies a transformation on the output variable was applied.
This transformed model with only two parameters performed in a superior way and was even able to predict colour development without the use of temperature and radiation data.
The models were fitted to an identification data set to identify the model parameters.
Then their prediction performance was evaluated using a validation data set. In a second step a harvest prediction model was developed using the fruit colour development model.
Given the colour stage distribution of the bell peppers at the time of prediction the amount of harvestable fruits in the next week or two weeks was predicted.
A root mean squared error of prediction (RMSEP) of 8.4% and 11.7% was achieved using the identification and validation data set respectively for a 14 day prediction.
A method was developed to predict bell pepper harvest with a time horizon of one week and two weeks on a greenhouse level.
In a first stage several fruit colour development models were proposed.
A first group consisted of ARX models taking into account the greenhouse temperature and solar radiation data.
Because of some model inadequacies a transformation on the output variable was applied.
This transformed model with only two parameters performed in a superior way and was even able to predict colour development without the use of temperature and radiation data.
The models were fitted to an identification data set to identify the model parameters.
Then their prediction performance was evaluated using a validation data set. In a second step a harvest prediction model was developed using the fruit colour development model.
Given the colour stage distribution of the bell peppers at the time of prediction the amount of harvestable fruits in the next week or two weeks was predicted.
A root mean squared error of prediction (RMSEP) of 8.4% and 11.7% was achieved using the identification and validation data set respectively for a 14 day prediction.
Authors
B.E. Verlinden, B.M. Nicolaï, C. Sauviller, W. Baets
Keywords
harvest prediction, glass house, Capsicum annuum, modelling
Online Articles (81)