Groundwater contamination sources identification based on kernel extreme learning machine and its effect due to...

Area of Science:

• Environmental Science
• Geochemistry
• Data Science

Background:

• Groundwater contamination source identification (GCSI) accuracy relies heavily on measurement data quality.
• Measurement noise from errors can significantly impact the reliability of GCSI results.

Purpose of the Study:

• To evaluate the effectiveness of wavelet hierarchical threshold denoising for contaminant concentration measurements.
• To assess the impact of denoised data on the accuracy of groundwater contamination source identification.
• To develop an efficient model for identifying contamination source location and release history.

Main Methods:

• Wavelet hierarchical threshold denoising applied to contaminant concentration data.
• Groundwater contamination sources identification (GCSI) using denoised measurements.
• A 0-1 mixed-integer nonlinear programming (0-1 MINLP) optimization model integrated with a kernel extreme learning machine (KELM) for source identification.

Main Results:

• Wavelet denoising was more effective with continuous (2-day) than discrete (2-month) measurements.
• GCSI accuracy improved significantly when using denoised data compared to noisy data.
• The KELM-based 0-1 MINLP model accurately identified source location and release history, reducing computation time by 96.5%.

Conclusions:

• Data denoising is crucial for accurate groundwater contamination source identification.
• The proposed KELM-based 0-1 MINLP model offers an efficient and accurate solution for GCSI.
• This approach enhances the reliability and efficiency of managing groundwater contamination.