Water quality early warning method based on Informer model

Abstract

The invention discloses a water quality early warning method based on an Informer model. The water quality early warning method comprises two parts of time sequence water quality data prediction and an anomaly detection model. The method comprises the following steps: firstly, selecting time sequence data, supplementing missing values, and carrying out polynomial smoothing filtering (Savitzky-Golay, SG) processing on the supplemented data; and on the basis, training the data after SG filtering and smoothing processing, then performing Loess-based seasonal characteristic decomposition on a predicted value output by the generated time sequence prediction model, and eliminating a seasonal component. Besides, modeling is carried out on the decomposed multi-dimensional feature data by adopting an isolated forest, and data exceeding a safety interval is set as an abnormal value, so that the effect of water quality early warning is achieved. Wherein the abnormal value is combined with the environmental quality standard for surface water GB3838-2002 to serve as a judgment standard. According to the method, the problems of real-time prediction and security interval setting can be effectively solved, so that the accuracy rate and the recall rate are further improved.

Description

Technical field [0001] The present invention relates to an early warning method for water quality indicators, mainly to design a water quality early warning method based on the Informer model. Background [0002] In recent years, with the development of Internet of Things technology, various detection data can be effectively transmitted and analyzed in a timely manner, while the acquisition of data has high frequency, and the scale of data volume is enough to meet the application of neural network methods. Time series refers to the numerical value of the same statistical indicator arranged in the order of its occurrence, there is a certain law to affect each other at adjacent times, and it is difficult for traditional neural networks to capture this long span of key information, resulting in a decrease in prediction accuracy. Although recurrent neural networks can be fed into the model in the order in which the data is trained, it is difficult to pay attention to two time series...

AI Technical Summary

Problems solved by technology

In the water quality early warning work, most of the traditional early warning methods set a fixed early warning threshold. If the threshold is set high, the sensitivity of the early warning system will be low, which will affect the emergency treatment of sudden pollution time; if the threshold is set too low, the early warning system will Frequent alarms increase unnecessary workload

Images