利用物聯網及普通克利金法進行溫室溫度感測器最佳化設計

Abstract

溫室農業在台灣非常普及，許多溫室納入物聯網監測系統以幫助溫室進行環境控制。然而在台灣，由於大多數農戶屬於中小型業者，過度的放置溫度感測器除了可能無益於作物的生長，同時會增加成本支出，在維護上以及農作上也倍具考驗。本研究將開發出一套低價的物聯網監測系統幫助降低監測成本，隨後利用普通克利金法決定溫室所需溫度感測器數量幫助農民決定合宜的感測器數量，奠定往後針對溫室感測器對於作物生長的影響的相關研究的基礎。本研究選用位於台北以及台中兩處溫室，於台北的溫室探討不同操作模式對於溫室所需感測器的影響，於台中的溫室則探討不同感測器初始位置對於所需感測器數量的影響。本研究共開發出三種決定感測器數量的策略。結果顯示策略一最嚴格，策略二次之，策略三最寬鬆。三者所代表的物理意義不同，農民可以基於自己的需求進行挑選。此外，不同環控情境下最適合用以回歸的變異函數亦不相同，以台北溫室的例子，在環控採自動控制時適合用gaussian變異函數，在負壓風扇開啟時適合用spherical變異函數，在負壓風扇及水簾同時開啟時適合用exponential變異函數，在台中的溫室無論感測器布置為何，最適合之變異函數均為exponential變異函數。研究中也發現，溫室內部平均溫度與溫室內部溫度分散程度呈現高度正相關。負壓風扇的開啟能夠幫助感測器數目減少，然而水簾的作用會抑制負壓風扇的作用，使得所需感測器數量變多。

Greenhouse farming is very ubiquitous in Taiwan, and many greenhouses are equipped with IoT monitoring systems to help environmental control. However, in Taiwan, since most farmers are small and medium-sized businesses, the excessive placement of temperature sensors may not only be benefitial to the growth of crops, but also increase costs, and it is also more challenging in maintenance and farming. This research first expects to develop a low-cost IoT monitoring system to help reduce monitoring costs, and then develop strategies based on ordinary kriging to determine the number of temperature sensors needed in a greenhouse to help farmers determine the appropriate number of sensors, laying the foundation for future research on the impact of greenhouse sensors on crop growth. Two greenhouses in Taipei and Taichung were selected for this study. The effect of different operation modes on the sensors required in the greenhouse was investigated in the greenhouse in Taipei, and the effect of different initial positions of the sensors on the number of sensors required in the greenhouse in Taichung was investigated. Three strategies for determining the number of sensors were developed in this study. The results show that strategy 1 is the strictest, strategy 2 is the second, and strategy 3 is the most lenient. The physical meanings represented by the three are different, and farmers can choose based on their own needs. In addition, the most suitable variogram for regression under different environmental control situations are different. Take the Taipei greenhouse as an example, the gaussian variogram is suitable when the environmental control adopts automatic control, and the spherical variogram is suitable when the fan is turned on. When the fan and the water curtain are turned on at the same time, the exponential variogram is suitable. When the fan and the water curtain are turned on at the same time, the exponential variogram is suitable. In the greenhouse in Taichung, no matter what the sensor arrangement is, the most suitable variogram is the exponential variogram. The study also found that the average temperature inside the greenhouse was highly positively correlated with the degree of temperature dispersion inside the greenhouse. Turning on fan can help reduce the number of sensors. However, the function of the water curtain will inhibit the function of the negative pressure fan, so that the number of required sensors is slightly increased.