A Multi-Source Fusion Groundwater Level Prediction System and Method Based on Dual-Model Machine Learning

By employing a dual-model machine learning architecture that combines random forest regression and Kriging interpolation models, the problems of spatiotemporal fragmentation and insufficient spatial optimization in groundwater level prediction are solved. This results in high-precision and practical groundwater level prediction, adapting to the changing characteristics of different water level indicators and improving the stability and scalability of the model.

CN122133873APending Publication Date: 2026-06-02POWERCHINA BEIJING ENG CORP
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
POWERCHINA BEIJING ENG CORP
Filing Date
2026-03-17
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing technologies for groundwater level prediction suffer from problems such as spatiotemporal fragmentation, lack of hydrological adaptability in spatial optimization, lack of in-depth optimization chain in prediction results, and insufficient targeting of dual-objective predictions. As a result, the prediction results are difficult to balance temporal continuity and spatial consistency, and have low engineering practicality, poor model stability and scalability.

Method used

A dual-model machine learning architecture is adopted, combining a random forest regression model and a kriging interpolation model. Through multi-dimensional data preprocessing, feature engineering, and weighted fusion, a prediction model with temporal and spatial dimensions is constructed, integrating multi-source data to generate the final groundwater level prediction result.

Benefits of technology

It achieves spatiotemporal collaborative prediction, improves the temporal continuity and spatial consistency of prediction results, enhances the adaptability to multi-source data, adapts to the changing characteristics of different water level indicators, lowers the technical implementation threshold, and improves the stability and scalability of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122133873A_ABST
    Figure CN122133873A_ABST
Patent Text Reader

Abstract

This invention, entitled "A Multi-Source Fusion Groundwater Level Prediction System and Method Based on Dual-Model Machine Learning," belongs to the field of groundwater level prediction technology. The technical problem it aims to solve is the disconnect between spatiotemporal analysis and time prediction in existing groundwater level prediction methods, insufficient multi-source data adaptation and processing capabilities, and a lack of specificity in predictions that deviate from hydrological patterns. The key technical solution is as follows: the system comprises six functional modules. First, multi-dimensional, multi-source data is collected and standardized preprocessed. Then, a feature set is constructed and optimized, and a random forest temporal dimension prediction model and a Kriging spatial dimension optimization model are trained. Initial predictions generate preliminary values, which are then validated. Next, spatial weighted fusion optimization is performed, and finally, after verifying and correcting outliers, predicted data, charts, and reports are output. The method simultaneously implements this entire process.
Need to check novelty before this filing date? Find Prior Art