Coal mine gas concentration prediction method combining ensemble learning and weighted extreme learning machine

Abstract

The invention discloses a coal mine gas concentration prediction method combining ensemble learning and a weighted extreme learning machine. The method comprises the steps: obtaining historical gas concentration monitoring data, carrying out the data preprocessing and missing data filling, generating a sample set, splitting a prediction target Y in the sample set phi, and enabling each column of data in the prediction target Y to be combined with X; according to the method, the latest gas concentration data can be received in real time, gas concentration trend prediction for the next three hours is formed, and the method is low in time complexity and high in accuracy.

Description

technical field [0001] The invention relates to a coal mine gas concentration prediction method, which belongs to the technical field of gas concentration prediction. Background technique [0002] Coal mine gas protection is a high-risk industry, which can easily cause mass deaths and injuries. According to statistics, gas explosion accidents account for more than 60% of major accidents in my country, and the prevention and control of gas accidents is the top priority of coal mine safety work. With the development of artificial intelligence technology, the intelligent prediction of gas concentration has become the main means of preventing gas explosion accidents. At present, gas concentration prediction methods mainly include 1) time series prediction, such as exponential smoothing and curve fitting, etc. This method has many parameters, low prediction accuracy, and difficult model update. 2) The bp neural network and other technologies are used for prediction. This kind o...

AI Technical Summary

Problems solved by technology

At present, the gas concentration prediction methods mainly include 1) time series prediction, such as exponential smoothing and curve fitting, etc., this method has many parameters, low prediction accuracy, and difficult to update the model

2) Using bp neural network and other technologies to make predictions, this kind of method takes a long time to train and the dynamic adjustment is more complicated

The above method only considers the case of a single classifier, and there are certain differences between single classifiers, which may lead to errors in classification

Images