A water quality prediction method based on k-nearest-neighbor probability rough sets and PSO-LSTM

Water security has attracted a lot of attention in the world, and water quality assessment is the main task to ensure water security. In China, as an important water supply route in the Beijing-Tianjin-Hebei region, the South-North Water Transfer Middle Route is crucial to the economic development and health of people in the region. Therefore, an effective water quality prediction method is essential to prevent water quality degradation and water pollution in the South-North Water Transfer Line. In this paper, on the basis of the water quality data of 13 automatic monitoring stations in the middle of the Transfer Middle Route, we propose a water quality prediction method based on k-nearest-neighbor probability rough sets and PSO-LSTM algorithm. More specifically, we first introduce a novel model of k-nearest-neighbor probabilistic rough sets (KNPRSs) by combining k-nearest-neighbor algorithm and probabilistic rough sets. Then, an attribute reduction approach based on KNPRSs is developed, which can effectively eliminate the redundant attributes in water quality assessment and filter out the valuable feature attributes. Furthermore, we propose a water quality prediction method based on LSTM neural network model optimized by PSO algorithm. By introducing the PSO algorithm, the hyperparameters of the LSTM neural network are adaptively optimized to improve the accuracy of water quality prediction. At last, the historical data of three automatic monitoring stations along the route are selected, and six water quality indicators with practical forecasting value are used as the target, and a comparison experiment is conducted. The experimental results show that the KNPRSs-PSO-LSTM model can fully extract the key characteristics of water quality attributes, can be used to predict different target indicators at different stations, which is reliable, stable and efficient, can effectively improve the prediction accuracy, can be applied to the South-North Water Transfer Middle Route water quality forecasting and early warning work.