스마트 팜에서 생장 단계별 손실함수를 고려한 인공신경망 기반 최적 환경 탐색 시스템 = An Artificial Neural Network–Based Optimal Environment Search System Considering Growth Stage–Specific Loss Functions in Smart Farms

다국어 초록 (Multilingual Abstract)

In this paper, we propose an artificial neural network–based optimal environment search system to optimize crop growth in smart farms using growth stage–specific loss functions. The system applies customized losses for vegetative, reproductive, and balanced stages, linking a growth prediction model with an optimal environment search model to derive stage-specific optimal conditions. It consists of data collection, preprocessing, merging, growth prediction, and environment search modules. The growth prediction model uses an RNN-based time-series structure to predict key indicators such as flower-cluster distance and stem diameter, while the optimal environment search model searches for the optimal combination of temperature, humidity, and CO₂ based on predicted growth states. In addition, environmental constraints ensure realistic variable ranges. Experimental results show that the system reproduces stable growth trends even with limited data and can identify realistic optimal growth environments without complex experiments or large-scale measurements.