數位濾波處理於水耕栽培即時離子感測之研究

Abstract

水耕種植中，植物生長所需營養素皆由水耕養液中獲得，但現今缺乏可即時量測離子濃度之儀器，導致植物可能在過多或缺乏某種離子下生長。實驗室先前透過平面固態式離子選擇電極與微控制整合，開發為即時離子量測系統，但量測濃度精準度受到雜訊干擾嚴重。因此本研究嘗試透過數位濾波器解決雜訊干擾問題，使用數位濾波器包含移動平均濾波器 (moving average filter, MAF)、卡爾曼濾波器 (Kalman filter, KF) 和中位數濾波器 (median filter, MF)。在植物工廠中以鉀、鈣和鎂離子選擇電極進行量測，並透過快速傅立葉轉換(fast Fourier transform, FFT)和實驗結果進行分析，探討數位濾波器對離子選擇電極量測系統影響和雜訊主要來源和頻率。 根據FFT分析發現系統所受主要雜訊頻率為60Hz，與台電公司所提供交流電頻率符合，且該雜訊對硬體和電極皆會造成干擾。透過鉀、鈣和鎂離子選擇電極實驗結果可證明，數位濾波器可提高電位訊號穩定性，以鉀離子選擇電極為例，電位訊號標準差可由13 mV降為1 mV以下，且並不影響電位與濃度線性迴歸。濃度精準度方面，透過一價鉀和二價鈣離子選擇電極比較可證明，精準度提高並不受價數影響，兩離子電極平均絕對誤差 (Mean absolute error , MAE) 皆可降為0.2 mM以下。而二價鎂離子選擇電極原始數據MAE為1.07 mM，但經過數位濾波器後MAE僅降低為0.73mM，與鈣離子選擇電極比較可證明，數位濾波器並無法改善靈敏度低所造成誤差。考量卡爾曼濾波器可以降低離群值影響，選擇卡爾曼濾波器作為數位濾波器應用，最後也在商業型水耕場域中進行實驗，量測鉀和鈣離子濃度與離子層析儀相比僅有2 mM誤差。 綜合以上，本研究證明數位濾波器可以提高離子選擇電極系統濃度精準度，且考量離群值對數位濾波器影響，選擇卡爾曼濾波器作為數位濾波器應用。未來可透過即時離子量測系統快速得到離子濃度資訊，並將養液中離子濃度維持在特定範圍內，即可針對植物生長時最佳離子濃度範圍進行研究。

The current monitoring and adjustment of hydroponic nutrient solution rely on conductivity and pH values, which do not provide information about the concentration of ions in the nutrient solution. Solid-contact ion-selective electrodes (SCISEs) can be integrated with microcontroller to create a real-time ion monitoring system (RTIMS). However, the measurement results are affected by noise, making it challenging to distinguish concentration changes accurately. This study employs digital filters, including moving average (MAF), Kalman (KF), and median filters (MF), to reduce the noise in the signal. In the plant factory, measurements are using potassium (K+), calcium (Ca2+), and magnesium (Mg2+), ion-selective electrodes. Using fast Fourier transform (FFT) to investigate the primary sources and frequencies of noise. Based on FFT analysis, it was found that the main noise frequency is 60Hz, which corresponds to the frequency of the AC power provided by Taiwan Power Company. This noise interferes with both hardware and electrodes. The experimental results of K+, Ca2+ and Mg2+ SCISEs demonstrate that digital filters can improve the stability of the potential signal. Taking K+ SCISE as an example, the standard deviation of the potential signal can be reduced from 13 mV to below 1 mV. The experimental results of K+ and Ca2+ SCISEs demonstrate that the accuracy is not influenced by the valency of ion. The mean absolute error (MAE) of K+ and Ca2+ SCISEs is below 0.2 mM. The original MAE of Mg2+ SCISE is 1.07 mM. However, the MAE is reduced to only 0.73 mM through the use of KF. Comparing with the Ca2+ SCISE, it can be demonstrated that the digital filters are unable to improve the performance of electrode. KF is chosen as the digital filter for RTIMS because it can reduce the impact of outliers. Finally, the RTIMS measures hydroponic nutrient solution in a commercial hydroponic green house. The concentration of K+ and Ca2+ determined by the RTIMS exhibit only 2 mM error when compared to the values obtained from the ion chromatograph. In summary, this study demonstrates that digital filters can improve the accuracy of the ion-selective electrode device. Furthermore, considering the impact of outliers on the digital filters, the KF is chosen as the digital filter for the RTIMS. In the future, the RTIMS can provide real-time ion concentration information, allowing for the maintenance of ion concentrations within specific ranges in the nutrient solution. This capability conduct research on the optimal ion concentration range for plant growth.