A method for detecting abnormal events in time series data of multivariable water quality parameters

Abstract

The invention discloses a method for detecting abnormal events of time series data of multivariable water quality parameters: first, input multiple water quality parameters to model, train and construct a data-driven prediction model (BP model), and analyze the time series of multivariable water quality in the water supply pipe network data, and evaluate the model; then, obtain the predicted value of water quality data through BP model prediction, compare the actual measured value of the current state with the predicted value obtained by using the prediction model for error evaluation and classification analysis, and determine the abnormal events of single variable parameters; based on Classify the error statistical results, determine the event probability of univariate water quality parameters through sequential update Bayesian update, and perform multivariate fusion decision-making, integrate information from multiple water quality monitoring indicators, provide unified decision results, and determine the water supply network Whether an abnormal event occurs at a specific node.

Description

technical field [0001] The invention relates to a method for detecting abnormal events of time series data of multivariable water quality parameters and the technical field of water supply network monitoring network. Background technique [0002] With the rapid economic development, abnormal water pollution incidents occur frequently. If a city's water supply system is polluted, it will bring huge losses to the society. Pollution events in the water distribution network spread rapidly and cause more damage. Therefore, timely detection of abnormal water quality events in the water supply network, timely warning of abnormal pollution events, and prevention of continued spread of pollution are of great practical significance. At present, wireless sensor networks are widely used in the monitoring of various scenarios. The parameters that can clearly reflect the water quality mainly include free chlorine, total organic carbon, electrical conductivity, pH value, temperature, and...

AI Technical Summary

Problems solved by technology

[0004] The above methods are all limited to judging abnormal water quality pollution events in the water supply pipe network based on whether a single water quality index exceeds the standard, and the single detection index does not conform to the real water supply pipe network water environment.

When a pollution event occurs in the water supply pipe, multiple parameters will change significantly, and the single factor considered for a single parameter is likely to cause missed and false positives, and the detection accuracy is not high

Images