A NEURAL NETWORK MODEL FOR MONITORING AND FORECASTING THE PROCESS OF STORING OILSEEDS
Abstract
The paper presents the results of research on the control, forecasting and management of the storage process of oilseeds. Vegetable oil production and energy consumption are two important factors of oilseed storage systems. The first ones should meet the needs of the population quantitatively and qualitatively. As a result, in order to properly control these two important aspects, it is necessary to maintain an appropriate microclimatic environment using a computational decision support system (DSS) that is able to adapt to changes in environmental characteristics. The multilayer perceptron neural network (MLP-NN) was developed to simulate the temperature and relative humidity profiles of the environment inside the storage. A specific NN uses Levenberg–Marquardt back propagation as a training algorithm; the input variables are external temperature and relative humidity, solar radiation, as well as internal temperature and relative humidity of the medium. The maximum errors in the calculations of temperature and relative humidity are 0.877 ℃ and 2.838%, respectively, while the determination coefficients are 0.999 for both parameters. A model with a low maximum error in forecasts will allow DSS to send appropriate control commands to the electrical drives of the storage to maintain internal conditions at the desired level for storage with the lowest possible energy consumption.