Methodology of development of intellectual energy efficient system of control of temperature-humidity regime in industrial heat

The cost of energy carriers affects the cost of products that are grown in industrial greenhouses. The purpose of the study is to develop a methodology for creating an intelligent system for energy-efficient microclimate management in industrial greenhouses operating under conditions of uncertainty, based on a combination of methods for predicting energy consumption and minimising them through the use of artificial neural networks. Methodology – the input parameters of the neural network prediction model are: the value of external and internal greenhouse air temperatures, the value of relative humidity, solar radiation absorbed by the greenhouse, and the level of carbon dioxide in the greenhouse. The outputs of the forecasting model are the values of natural gas and electricity costs. Based on the use of fuzzy logic methods and a genetic algorithm, models for finding and using optimal parameters of PI controller settings adapted to changes in the operating conditions of the automation system are developed and investigated. Methods of improving the quality of regulation of technological parameters by combining various intelligent control algorithms in one automation system are analysed, which helps to reduce energy costs by 10-13%. It is established that for closed-ground structures, heating and ventilation systems have the highest energy consumption (on average, more than 4,000 m3 of natural gas and almost 1,000 kWh of electricity are consumed per day for heating and ventilation in an industrial greenhouse. Correlation analysis of the relationships between external disturbances and energy costs that ensure compliance with a given technology for growing plant products, confirmed the hypothesis about the presence of conditions of uncertainty in the functioning of an industrial greenhouse formed by random disturbances, incomplete information about the biological component; while the linear correlation coefficients do not exceed r<0.35. This creates conditions for the use of neural networks both for predicting energy costs and for forming energy efficient management strategies. This provides better regulation in conditions of uncertainty, the adjustment time and overshoot are reduced by two to three times. To create an energy-efficient microclimate management system in industrial greenhouses that operates in conditions of uncertainty, a neural network model for predicting the energy costs of natural gas and electricity has been developed. The authors of the study have improved the structural and functional scheme of the automation system for controlling the temperature and humidity regime in industrial greenhouses by combining intelligent algorithms for stabilising the operation of technological equipment at the lower level of control and optimising energy costs by predicting them at the upper level. The introduction of such a system allows to saving natural gas for heating up to 13% and electricity – up to 10%. The study proved that the development of an intelligent energy-efficient system for automatic control of microclimate parameters in closed-ground structures should be based on a combination of methods for intelligent adjustment of PI controller parameters and forecasting energy consumption.