Experimental model and neural network based electrical conductivity estimation in soilless culture system

Area of Science:

• Agricultural Science
• Hydroponics
• Nutrient Management

Background:

• Effective nutrient solution management is crucial for soilless culture systems, particularly in recirculating hydroponics.
• Accurate prediction of electrical conductivity (EC) is necessary to maintain optimal nutrient levels after fertilizer application.

Purpose of the Study:

• To develop theoretical and artificial intelligence-based models for predicting the EC of complex nutrient solutions in soilless culture.
• To extend existing EC prediction methods for multi-component electrolyte solutions.

Main Methods:

• Developed an extended theoretical EC prediction equation based on Robinson and Stroke's model.
• Implemented a three-layer multilayer perceptron neural network with backpropagation for EC prediction.
• Utilized concentrations of seven macro elements, Na+, and Cl as input variables for the models.

Main Results:

• Both the experimental EC prediction equation and the neural network model demonstrated good agreement with measured EC values.
• The developed models accurately predicted the electrical conductivity of nutrient solutions containing multiple inorganic compounds.
• The study validated the efficacy of both theoretical extension and machine learning approaches for EC prediction.

Conclusions:

• The developed EC prediction models are effective for managing nutrient solutions in soilless culture.
• These models can aid in optimizing nutrient delivery and maintaining solution stability in hydroponic systems.
• Accurate EC prediction is essential for efficient and sustainable soilless cultivation practices.