87 / 2021-09-26 18:13:59

Research on intelligent prediction model of gas concentration in working face based on CS-LSTM

gas concentration,intelligent prediction,CS,LSTM,optimization

The monitoring of gas concentration in working face is one of the main means to prevent and control gas accidents in coal mine. The research on the prediction model of gas concentration in working face can provide a certain basis for realizing the intelligent sensing of gas disaster in coal mine. In order to fully excavate the monitoring information of gas concentration in working face, a CS-LSTM intelligent prediction model for gas concentration in working face was proposed based on the advantage of Long Short-Term Memory (LSTM) model for time series data processing and the global optimization ability of Cuckoo Search (CS) algorithm. The model takes the monitoring data of gas concentration in working face as the sample. Firstly, spline interpolation method is used to interpolate the missing values in the samples, and the maximum and minimum value method is used to conduct dimensionless processing of the data to get the training samples. Second the mean square error (MSE) was used as the evaluation index of LSTM prediction model, and the reciprocal of MSE was used as the fitness function of CS algorithm. Then, the global optimization performance of CS algorithm was used to optimize the four super parameters of LSTM model, including the number of hidden layers, neurons in hidden layers, the number of fully connected layers and neurons in fully connected layers. The grid structure of LSTM model which is most suitable for gas concentration prediction of working face is obtained by outputting the optimal combination of super parameters. Finally, the optimal combination of super parameters is given to the LSTM model, and the optimal intelligent prediction model of gas concentration is constructed and verified by an example. The results showed that: Under the same number of iterations, CS algorithm has a wider search range, which effectively avoids the shortcoming of genetic algorithm (GA) that is easy to fall into local optimal, and has a better global optimization ability. Compared with LSTM prediction model and GA-LSTM prediction model, the MSE of CS-LSTM prediction model is 0.000546, which has the best effect and the highest accuracy. It can meet the prediction requirement of gas concentration in working face.