Underground water level missing data restoration method based on geostatistics and neural network

Abstract

The invention relates to the technical field of underground water level restoration, in particular to an underground water level missing data restoration method based on geostatistics and a neural network. According to the invention, a restoration model is constructed from the aspects of space and time, and the accuracy and reliability of data missing interpolation are improved. The method comprises the following steps: dividing a spatio-temporal data set of an underground water level into a spatio-temporal sequence data set and a time sequence data set; according to the time-space sequence data set, adopting a Kriging method to interpolate missing values to carry out spatial interpolation based on geostatistics, and obtaining a restored spatial data set; establishing a time sequence missing value restoration model based on a BP neural network, and restoring missing data of the time sequence data set; carrying out training and learning according to the space restoration result, the time restoration result and the actual detection data, continuously adjusting the space-time element weight through machine learning, and constructing a mixed restoration model of the underground water level space-time sequence missing data; and repairing missing data in the underground water level space-time monitoring data set by utilizing the hybrid repair model.

Description

technical field [0001] The invention relates to the technical field of groundwater level restoration, in particular to a method for restoring groundwater level missing data based on geostatistics and neural networks. Background technique [0002] Among the missing data repair methods based on statistics, the earliest methods used to fill missing data include mean filling method, hierarchical mean filling method, maximum expectation algorithm, multiple filling method and other statistical methods. Among them, the mean filling method uses the mean value of the two observations before and after or the mean value of the non-missing data to fill the missing value; the layered mean filling method stratifies the groundwater data according to the same month of each year, and takes the mean value of each layer instead of Missing values; the maximum expectation algorithm uses mathematical formulas to estimate the maximum likelihood of missing data, and iterates continuously, which has...

AI Technical Summary

Problems solved by technology

Aiming at the spatio-temporal characteristics of groundwater level monitoring values, this method uses geostatistics theory and artificial neural network algorithm to construct a spatial restoration model and a time restoration model suitable for groundwater level monitoring value characteristics from the perspective of space and time, and expands the method according to the spatiotemporal elements , to build a hybrid model that can better evaluate the spatio-temporal elements and have higher precision for missing groundwater level data repair, which improves the accuracy and reliability of data missing interpolation

Images