使用機器學習技術預測溫室室內環境以實現溫室環境之自動控制

Predict indoor environment of greenhouses for automatic greenhouse environmental control using machine learning techniques

徐嘉徽(Chia-Hui Hsu)

指導教授 ： 張斐章

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202002954

摘要

維持穩定之作物產量是溫室耕種的主要優勢，然而為控制溫室內部環境，相較於傳統的露地栽培亦會消耗額外的資源；以水-能源-糧食鏈結(Water-Energy-Food Nexus, WEF Nexus)之管理為出發點，首要工作是建立一套能評估並最佳化溫室作物產量及資源消耗之整合性方法；由於溫室之作物產量在其內部環境受到妥善控制的前提下是可預期的，如何盡可能地減少控制過程中所消耗之水資源及能源是主要的探討重點。

為達成此目的，本次研究首先建立了機器學習模式以預測種植溫室之內部環境，包括氣溫、相對濕度，以及土壤含水量；並根據適宜作物生長之環境標準，若判斷預測結果超出此一標準，則會啟動環境之控制機制。針對判斷需啟動控制機制的情況，本次研究另外建立一評估模式決定適合之控制方式及其控制幅度，以調節溫室環境至符合環境標準之範圍。本次之研究區域為一位於彰化縣，種植小黃瓜及番茄之溫室，通過物聯網設備收集2019/01/08至2019/11/12之環境監測資料，資料解析度為10分鐘一筆，共計44304筆資料。針對溫室氣溫的部分，以30℃作為上限，並以外遮蔭作為控制手段時，結果顯示相較於「未來第一小時之氣溫預測值在30℃以上」，使用「當時刻氣溫觀測值在27℃以上」作為操作標準在低程度，即遮蔽面積較少之方案的選擇次數上明顯較前者多，顯示通過預測模式提前對未來之高溫情況作出應對能有效提升溫室控制內部氣溫之效率。

關鍵字

水-能源-糧食鏈結 ； 溫室 ； 機器學習 ； 物聯網

並列摘要

Maintaining stable crop production is the main advantage of greenhouses. However, it would also consume additional resources to control its indoor environment, as compared to traditional open-field cultivation. In consideration of Water-Food-Energy Nexus (WFE Nexus) management, it’s essential to build an integrated methodology to estimate and optimize the production efficiency of greenhouses. Since the production of greenhouses is predictable if the indoor environment is well controlled, the main issue we should think about is how to reduce the consumption of resources as much as possible during the control process for greenhouse indoor environment.

For this purpose, we first build a machine learning-based model to predict indoor environment condition, including air temperature, relative humidity (RH), and soil volume water content (VWC), for the study greenhouse. Then according to the suitability criteria of the crop, the predicted values are utilized for environmental control if the values violate the requirements. Under such conditions, an estimation model is built to decide appropriate type and degree of control mechanisms for meeting the criteria to maintain stable crop production. The study area is a greenhouse located at the farm in Changhua County, Taiwan, cultivating cucumber and tomato. A total of 44,304 datasets were recorded by Internet of Things (IoT) from 2019/01/08 to 2019/11/12 at a 10-minute temporal resolution. For greenhouse indoor temperature control, we use 30℃ as the upper limit and shade cloth as the control method. The results show the operation efficiency is better with the prediction model than just utilizing real-time observation value for environmental control.

並列關鍵字

Water-Food-Energy Nexus (WFE Nexus) ； Greenhouse ； Machine learning ； Internet of Things (IoT)

參考文獻

1. Al-Mezeini, N. K., Oukil, A., Al-Ismaili, A. M. (2020). Investigating the efficiency of greenhouse production in Oman: A two-stage approach based on Data Envelopment Analysis and double bootstrapping. Journal of Cleaner Production, 247, 119160.

2. Ammar, M. E., Davies, E. G. (2019). On the accuracy of crop production and water requirement calculations: Process-based crop modeling at daily, semi-weekly, and weekly time steps for integrated assessments. Journal of environmental management, 238, 460-472.

3. Chang, L. C., Shen, H. Y., Chang, F. J. (2014). Regional flood inundation nowcast using hybrid SOM and dynamic neural networks. Journal of Hydrology, 519, 476-489.

4. Chen, J., Yang, J., Zhao, J., Xu, F., Shen, Z., Zhang, L. (2016). Energy demand forecasting of the greenhouses using nonlinear models based on model optimized prediction method. Neurocomputing, 174, 1087-1100.

5. Choab, N., Allouhi, A., El Maakoul, A., Kousksou, T., Saadeddine, S., Jamil, A. (2019). Review on greenhouse microclimate and application: Design parameters, thermal modeling and simulation, climate controlling technologies. Solar Energy, 191, 109-137.