經驗作物模式與滾動修正模組於溫室葉菜類蔬菜之生產排程系統

Abstract

葉菜類在臺灣等亞熱帶地區種植易受天候變化影響，不易保存且存放時間過長會導致品質下降。依據規格化訂單需求研擬生產計畫將有助於臺灣農民提升葉菜生產效率並降低成本。本研究目標是結合溫室物聯網設備量測的環境數據，建立葉菜類蔬菜產銷訂單需求排程的統計模型，根據出貨規格與預期收穫日進行生產排程，其中預期收穫日是以訂單預計出貨日期及可容許的冷藏天數估算。本研究對象為設施短期不結球型葉菜類: 青梗白菜 (Brassica chinensis L. cv. Ching-Geeng) 與小白菜 (Brassica chinensis L.)。開發該兩種葉菜類作物生長進行統計經驗模式，以光與溫度預估鮮食重量，藉由經驗作物模式獲得預期收穫日以推得預計種植的日期。由於實際作物生長有跨年間、跨季度、跨田區的差異，我們透過開發滾動修正模組，將真實田間作物生長資訊與經驗模式所模擬之作物生長預測結合，達到準確生產排程的目的。原始輻射量經驗模式經過滾動修正模組後，均方根誤差有明顯下降，表示預測值與真實值的平均偏離程度有因為滾動修正而降低，對模擬作物鮮食重量的精確度提升，所預計估的收穫日更貼近實際收穫日。這項研究結果提供農民在生產排程中，藉由滾動修正即時更新作物實際生長狀況，提升農業經營效益。

Leafy green vegetables grown in subtropical regions such as Taiwan are susceptible to weather changes, making them difficult to store and leading to deterioration in quality over long periods of time. Therefore, developing a production plan based on demands would be one way to improve production efficiency and reduce the costs of leafy vegetables. The objective of this study was to develop a protocol for scheduling leafy vegetable production using environmental data measured by IoT devices. The production schedule aimed to best satisfy both the delivery specifications (including plant length and fresh weight) and the expected delivery date of the order with some allowance for storage in the refrigerator. The crops of interest in this study included Brassica chinensis L. CV. Ching-Geeng, Brassica rapa L. ssp. chinensis Jusl. and Ipomoea aquatica. Firstly, the plant length and fresh weight of the above three leafy vegetables were estimated by the cumulative radiation exponential function. When there was a discrepancy between the model prediction and the natural growth, the crop growth simulation results were real-time corrected by modifying the model parameters according to the actual sampling results. After applying the real-time Adjustment Module to the original radiation-based empirical model, there was a significant decrease in Root Mean Square Error, indicating a reduction in the average deviation between predicted values and actual values. This reduction in error demonstrates an improved accuracy in simulating crop fresh weight. Additionally, the estimated harvest date is expected to align more closely with the actual harvest date. In conclusion, we believe the results of this study could provide farmers with the opportunity to update the actual growing conditions in the field to improve the efficiency of their agricultural operations.