Water quality index prediction method based on hybrid long-short-term memory neural network

Abstract

The invention discloses a water quality index prediction method based on a hybrid long-short-term memory neural network. First, the acquired historical data of the water quality index is sorted according to time series, and the historical water quality data is preprocessed by SG filter smoothing. Then, normalize the water quality data, divide the water quality time series data into multiple subsequences according to the preset sliding window size as the feature sequence, that is, after converting it into supervised data, input the data based on the encoder-decoder The long-short-term memory ED‑LSTM neural network model predicts the values ​​of water quality indicators at multiple time points in the future, and finally obtains the prediction results of water quality indicators with high accuracy.

Description

technical field [0001] The invention belongs to the technical field of water quality index prediction, in particular to a water quality index prediction method based on a mixed long-short-term memory neural network. Background technique [0002] Water quality indicators can be used as a specific measure to judge the degree of water pollution. Real-time collection and acquisition of water quality index data by surface water quality automatic monitoring stations. Since the obtained data changes dynamically with time, the time series analysis and prediction of its historical data can be used to know the change trend of water quality, and then provide support for water resource management and decision-making. Water environment indicators are affected by many complex factors such as physics, chemistry, and biology, and have strong nonlinear characteristics, without specific periodicity and stability. Traditional water quality index prediction methods generally use linear models...

AI Technical Summary

Problems solved by technology

At present, most water quality index data belong to long-term correlation time series, that is, there may be important events with relatively long intervals or delays in the series, but they have a greater impact on the next time value. It is difficult for traditional neural networks to capture such long-span events. Key information, leading to insufficient prediction accuracy

Images