Using Historical Data in the NNARX Model to Improve the Accuracy of Microclimate Parameter Forecasting

Abstract

The paper considers an approach to forecasting microclimate parameters in industrial premises based on a neural network of the NNARX type using historical data. The feasibility of incorporating time delays of input and output parameters is substantiated, taking into account the inertial nature of microclimate processes. The role of the autoregressive component in forming the internal “memory” of the model and ensuring temporal consistency of forecasts is demonstrated. The proposed approach establishes methodological prerequisites for the application of NNARX in intelligent microclimate control systems operating under dynamic industrial conditions.