Design Of Quantum Neural Network And Its Application In The Evaluation Of Super Deep Reservoir

This thesis studies the structure and algorithm design of the quantum-inspired BP network model,as well as the specific application of deep and ultra-deep reservoir evaluation in oil and gas fields.Based on the study of biological neurons and artificial neurons,a quantum neuron model based on quantum gates is proposed.This neural model is integrated into the BP network to construct a new quantum-inspired BP network model.In the topological structure of the model,quantum neurons are used in the input layer and the hidden layer,and ordinary neurons are used in the output layer.The input data of the network is first converted into a quantum state,and then aggregated under the action of hidden layer controlled non-gates and revolving gates.The final network output is generated under the action of ordinary neurons in the output layer.Experimental results comparing with ordinary BP network show that the model has improved the nonlinear mapping ability,robustness,convergence speed and rate of the network model after the introduction of the quantum computer system.In order to optimize the training process of the network model,an improved particle swarm optimization algorithm is proposed.The algorithm uses a Beta distribution method to initialize the population,adjusts the inertia factor using the inverse incomplete gamma function,and introduces a new equilibrium operator to In the speed update type,the problem of particle cross-border is solved by the symmetrical mapping method.The improvement measures proposed above achieve a balance between the exploration and development of the algorithm and improve the global optimization ability of the algorithm.The algorithm is applied in the training process of quantum-inspired BP network,which makes up for the shortcomings of traditional gradient descent methods such as slow convergence speed,and obviously improves the convergence speed and mapping ability of network training process.Finally,by comprehensively analyzing the well logging data of the oil and gas fields,the deep,super-deep reservoir evaluation index set is constructed,which has the characteristics of multiple input indicators and large sample data volume.Compared with the traditional neural network model algorithm,it is found that when the network has many input indicators and a large amount of network data,this model still has strong ability to approach and map complex relationships to identify reservoirs.The quantum-inspired BP network model designed in this thesis has strong adaptability and high prediction ability for the deep and ultra-deep reservoir identification problems in complex geological environments.