딥러닝 기반 시설재배 생산량 예측 연구

Smart agriculture optimizes labor and resources and increases production efficiency through data-driven decision-making. The field of predicting production based on data collected in real time is an important technology for productivity improvement and automation. In this paper, a deep learning model for forecasting production is used using environmental and growth information of paprika and cucumber among facility horticulture smart farms. By applying MLP(Multi Layer Perceptron), RNN(Recurrent Neural Networks), LSTM(Long Short-Term Memory models), GRU(Gated Recurrent Unit), TCN(Temporal Convolution Network), Transformer, etc., which are mainly used for time-series forecasting, the size of the look-back window and the forecasted data size Various adjustments were made to analyze the results of forecasting performance according to the model. The data of the smartfarm datamart was used, and the relationship between the size of the look-back window and the size of the forecasted to be predicted was investigated for each model.