Research On LLTSA Method Based On Nonlinear Data Dimensionality Reduction

In the industrial production process,water supply is needed,and the daily life of urban residents is inseparable from water.As an indispensable resource in our lives,water has not been protected and valued by people for a long time.Due to the inadequate sewage treatment and the inadequate discharge standards,it flows in.Rivers have led to the emergence of water pollution events in the surrounding areas of some areas,which has caused further deterioration of the water source.How to monitor and analyze complex water quality and quickly and effectively establish mathematical models are essential to improve wastewater treatment capacity.The complexity of sewage water quality is affected by many environmental factors,and the monitoring samples have the disadvantage of redundancy.They are nonlinear,time-varying,and have different main impact indicators for the water quality of different pollution sources.It is difficult to establish an analytical model intuitively and accurately.In this case Conduct research on sewage samples,simplify the input variables of the model,introduce data dimensionality reduction methods,and simplify low-dimensional variables with high-dimensional variables.As a current data dimensionality reduction method,the manifold learning algorithm has been concerned by a large number of scholars so far,and many new research results have been achieved,including linear algorithms and nonlinear algorithms.However,most of the existing data is mainly nonlinear when it becomes dominant,such as Sewage monitoring data and different nonlinear algorithms have different effects on processing data.The relevant theories of manifold learning and classic algorithms are introduced in detail,which theoretically supports the applicability of manifold learning methods in the reduction of sewage water quality dimensions.By studying and comparing these algorithms,we find the algorithm LLTSA used in this paper.It is superior to other general algorithms in processing data.It is locally linearized on the basis of tangent space arrangement,but there is a problem that the neighborhood parameters are difficult to determine.Use the neighborhood It is optimized by an adaptive method.The optimized tangent space can select different parameters in different neighborhoods.When used for sewage samples,it can retain the characteristics of the original data to a large extent and reduce the dimension of the data samples.In the establishment of the water quality prediction model,the variables obtained by high-dimensional sample compression processing are used as the input of the model to simplify the complexity of training the network.The water quality prediction model of BP neural network is established by preprocessing the data collected by the urban sewage plant.The wastewater treatment process and water quality parameters were thoroughly explored,the variables affecting the water quality were analyzed,the NALLTSA algorithm was used for processing,and the BPNN prediction model of water quality was established in combination with the BP neural network.The simulation experiment of the sewage plant data is compared with the traditional multiinput BP neural network prediction model.The prediction results show that the BPNN model achieves good prediction accuracy and can effectively save the time required for network training.This method has certain positive significance for sewage treatment and provides a reference for the prediction method of water quality.