Reservoir level dynamic monitoring method based on machine learning

By constructing a three-dimensional spatial coordinate system to screen suspicious environmental time intervals, identifying the fluctuation characteristics of reservoir water level data, and dynamically adjusting the learning rate of the GBDT model, the overfitting problem caused by a fixed learning rate is solved, thereby improving the prediction accuracy and reliability of dynamic monitoring of reservoir water levels.

CN122132884APending Publication Date: 2026-06-02河南省水利勘测设计研究有限公司
0 Cites 0 Cited by

Patent Information

Application Number
CN202610245663.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-02
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing gradient boosting decision tree (GBDT) models are prone to overfitting during the climate transition period when the fixed learning rate is used in reservoir water level dynamic monitoring, resulting in decreased prediction accuracy and an inability to effectively cope with noise interference during the seasonal climate transition phase.

Method used

By constructing a three-dimensional spatial coordinate system, screening suspicious environmental time intervals, identifying the fluctuation characteristics of multi-dimensional monitoring data, and dynamically adjusting the learning rate of the GBDT model, the model fitting range can be adaptively adjusted to improve prediction accuracy.

Benefits of technology

In complex scenarios such as climate transition periods, it improves the accuracy and reliability of reservoir water level dynamic monitoring, enabling efficient monitoring and early warning.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of data processing, in particular to a reservoir level dynamic monitoring method based on machine learning, and the method comprises the steps: obtaining multi-dimensional monitoring data of a reservoir at each monitoring moment in a preset historical time period before the current monitoring moment, constructing a three-dimensional space coordinate system, fitting data points in the three-dimensional space coordinate system, and obtaining a three-dimensional space coordinate system; obtaining a fitting straight line, and screening at least one suspicious environment time interval in a preset historical time period according to the distance difference between each data point in the three-dimensional space coordinate system and the fitting straight line; according to the fluctuation characteristics of the multi-dimensional monitoring data in each suspicious environment time interval, classifying the multi-dimensional monitoring data in each suspicious environment time interval to obtain a data classification result corresponding to each suspicious environment time interval, and adjusting a preset reference learning rate in a GBDT model to obtain an adaptive learning rate, based on the adaptive learning rate, the optimization model is obtained, and the accuracy of dynamic monitoring of the reservoir level is improved.
Need to check novelty before this filing date? Find Prior Art