Water quality evolution mechanism modeling and health risk assessment based on stochastic hybrid dynamic systems

Water quality assessment analysis is an important technical means for water pollution prevention and control. In this research area, mechanism models and real observation data of water quality evolution are always used to perform water quality assessment. However, the existing water quality evolution mechanism modeling researches commonly use a single time-invariant model to model the water quality evolution process. It is inappropriate to directly describe the complex behavior of long-term water quality evolution with the existing models, since the evolution process contain different feature states, and the water quality evolution characteristics under these states are different. In addition, the existing water quality assessment methods are mostly methods for directly processing and calculating the observation data of water quality. This makes the existing methods difficult to effectively compensate for the contingency and randomness in the water quality evolution process, which leads to deviations and errors in performing the water quality assessment. Considering these deficiencies, this paper proposes a water quality evolution mechanism modeling and health risk assessment method based on stochastic hybrid dynamic systems (SHDS). Firstly, a hybrid water quality evolution mechanism (H-WQEM) model is established based on SHDS, and a hybrid improved fruit fly optimization algorithm (H-IFFOA) is proposed to identify the unknown parameters of the H-WQEM model. Then, an improved interacting multiple model extended Kalman filter algorithm (IIMM-EKF) is employed to estimate the probability distribution of the hybrid state of the H-WQEM model, including the probability distribution of different feature states of water bodies and the probability distribution of water quality indexes under these states. Finally, the health degree of water quality is proposed as an indicator to achieve a quantitative assessment of the water health risk status. Real observation data from a monitoring station at Baiyangwan in China is used to validate the effectiveness of the proposed method. The results show that the method can effectively describe the complex water quality evolution process, and reasonably assess the water quality health risk status.