Study On CRHJ Neural Network Model In Mines Water Source Identification Based On Improved WOA Algorithm

It is very important to reduce the occurrence of mine water accidents for coal mine safety production.However,due to the complex hydrogeology condition of the mining area,it is difficult to prevent mine water disasters.Effective mine water source identification can not only prevent the occurrence of water disasters,but also locate the occurrence place quickly when water disasters occur,which is convenient for coal mining enterprises to take countermeasures and plays an essential part in the prevention of mine water disasters.Therefore,the key to prevent mine water disasters is to establish a reasonable mine water source discrimination model.Neural network is widely used to solve classification problems because of its strong nonlinear fitting ability.Cycle Reservoir with Hierarchical Jumps neural network(CRHJ)is used as the basic model in mine water source classification.In order to improve the accuracy of the model,Mixed Strategy based Improved Whale Optimization Algorithm(MSWOA)is proposed to optimize the weight parameters of CRHJ network,then improved CRHJ network is applied to build MSWOA-CRHJ mine water source discrimination model.The main work of this paper is as follows:(1)Aiming at the defects of Whale Optimization Algorithm(WOA),such as easily falling into local extremum and slow convergence speed,a Mixed Strategy Improved Whale Optimization Algorithm(MSWOA)is proposed.The performance of WOA has improved in four aspects:in order to solve the problem of slow convergence speed,low discrepancy sequence is used to initialize the population,which can improve the coverage of initial solution in solution space and the convergence speed of the algorithm;in order to solve the problem of poor balance between exploration and mining,nonlinear time-varying factor and adaptive weight strategy are proposed,which can improve the adjustment ability of the algorithm for overall search and local development;in order to solve the problem of incomplete development,the random learning strategy is used to increase the diversity of the population and improve the overall search ability of the algorithm;in order to solve the problem of easily falling into local extremum,Cauchy mutation operator is used to help the algorithm to jump out of local extremum.The experimental results on 12 benchmark functions and one practical engineering problem show that the MSWOA has significantly improved the optimization accuracy and convergence speed of WO A,which verifies the effectiveness of MSWOA.(2)Aiming at the problem of CRHJ network,which has many weight parameters and random values that affect the network performance,the MSWOA algorithm is used to optimize the performance of the CRHJ network,and the MSWOA-CRHJ algorithm is proposed.The algorithm is applied to mine water source identification,and the MSWOA-CRHJ mine water source identification model is established.In the process of mine water source identification,principal component analysis is used to reduce dimension and denoise original water quality data,the error function of the identification result is taken as the fitness function,and the reconstructed principal component data has inputted to train the model,and the final stage model output the optimal identification result of the sample water source.The MSWOA-CRHJ mine water source discrimination model has applied to the Dongtan Mine in Shandong Province.The result shows the identification accuracy of MSWOA-CRHJ model is 100%,which is higher than 90.909%of PCA-CRHJ,81.818%of CRHJ,81.818%of CRJ and 72.727%of ESN.It is proved that the model can accurately identify the type of water source and has higher identification accuracy,which verifies the reliability of the proposed model.