A multivariate based water mixing event detection method and performance comparison

The problems of internal and external water seepage in long-distance enclosed water conveyance project can cause potential engineering and water quality hazards. How to monitor and warn water mixing events quickly and accurately is a new topic. Conventional detection algorithms are often criticized for having high false positive rates and low true positive rates. This paper presents a new water mixed detection method based on multi parameter matching algorithm, which integrates Pearson correlation coefficient and Mahalanobis distance for water identification. The performance of the proposed method was evaluated using data from a water mixed experiment and compared Pearson correlation coefficient Euclidean distance-based method (PE) with Pearson correlation coefficient Mahalanobis distance-based method (PM). For the mixed events with large differences in water quality, both PE and PM methods can correctly detect 100% of all water mixing events with a 0% false alarm rate. For the mixed events with small differences in water quality, both PE and PM methods can correctly detect 95% of all water mixing events, and the false alarm rate are 1.92% and 0% respectively. Therefore, compared with the PE method, the PM method is less affected by the difference between external water and water in the main canal, has a larger optimal threshold selection range. The PM algorithm has higher detection performance and lower false alarm rate，which has better stability and can better distinguish between water quality fluctuations and fluctuations caused by mixing.