Research On Gas Emission Prediction Based On CIGOA-ENN Coupled Algorithm

The coal mine gas disaster is one of the major natural disasters that restrict mineproduction safety. It has seriously affected the lives and property safety of the workers,therefore, understanding the rules and characteristics of the gas disasters and achievingan accurate prediction of gas emission is an effective means to prevent gas disasters.The coal mine gas system is a complex, time-varying nonlinear dynamic system, sothe use of traditional linear prediction method to establish prediction model is diffcult tomeet the actual needs of the project. Therefore, this paper presents a new kind ofintelligent optimization algorithm-chaotic immune genetic optimization algorithm. ThenCIGOA is merged with Elman neural network to construct a non-linear dynamic systemidentification model based on CIGOA-ENN coupled algorithm for the prediction of gasemission. The model provides decision-making services for the prevention and treatmentof gas disasters.Firstly, this paper analyses the performance of the Elman neural network and itsadvantages and disadvantages. For the slow convergence, easy-to-precocious of theneural network, the paper proposed the use of genetic algorithm to improve it. Thenpaper analyses the operation mechanism of the genetic algorithm.When GA is used insolving multimodal, high-dimensional, nonlinear optimization problems, it is easy to fallinto premature convergence, so the paper presents chaotic immune genetic optimizationalgorithm by introducing the artificial immune thinking and chaos optimization idea toimprove GA. In the process of population evolutionary, the algorithm makes the particleclonal expansion, chaotic mutation, immune selection to improve the global searchability of the algorithm and enhance the accuracy of its search. Then CIGOA is mergedwith Elman neural network to establish nonlinear dynamic system identification modelbased on CIGOA-ENN coupled algorithm by using CIGOA to train Elman neuralnetwork weights and threshold. Using the historical data of mineactual monitoring to thesimulation, the results show that the model has faster convergence speed, higherconvergence accuracy and stronger robustness compared with the Elman and BP neuralnetwork models. It has provided a effective method to resolve the problem of coal minegas emission prediction.