FORMALIZATION OF MOISTURE TRANSFER IN AN UNSATURATED ZONE OF A MODULAR PLOT OF SOIL AS A CONTROL OBJECT BASED ON NEO-FUZZY NETWORKS

Abstract

The purpose of the article is to study a modular plot of soil as a control object, which is a complex system distributed in space. One of the characteristic features of the complexity of the control object is the uncertainty in representing its structure and behavior .Within the framework of the modern system modeling methodology, uncertainty may characterize the following aspects of model representations: vagueness or obscurity of the system boundary; ambiguity of separate terms’ semantics; incompleteness of model ideas about a certain complex system; the presence of contradictions between the individual components of the model representations or requirements that the model of a complex system must satisfy; the uncertainty of certain events’ occurrence relating to the possibility of finding the system - the original in a certain state in the future; linguistic uncertainty. The modular plot of soil as the object of control is influenced by variable disturbances - weather conditions (air temperature and humidity, wind speed, solar radiation, precipitation). The output parameter depends on them - the suction pressure (humidity) of soil. This object of control has stochastic uncertainty, since its properties change randomly. At present, in addition to classical neural networks, hybrid, in particular, fuzzy neural networks are developing. They combine the advantages of fuzzy inference neural networks. Therefore, it is the hybrid neuro-fuzzy network based on neo-fuzzy neuron that was used to model the unsaturated zone of the modular plot of soil as a control object. The developed neo-fuzzy models of unsaturated zone of a modular plot of soil as a control object for predicting the intake pressure of the soil provide higher accuracy of operation than multilayer direct distribution networks. At the same time, they have a simpler architecture, ensuring easier practical implementation and a higher learning rate. The neo-fuzzy models can be used as part of an automated workplace for an irrigation system dispatcher and serve as a convenient tool for planning and managing crop moisture regimes.