Parameter Optimization Of Doubly-Fed Induction Genrotor With Deep Learnning Based Quantum Parallel Multi-Layer Monte Carlo Algorithm

In a doubly-fed induction generator-based wind power generation system,with optimized parameters of the rotor-side controller(RSC)of this system,the control performance of the RSC can be improved,thereby further improving the wind energy conversion efficiency of this system.In recent years,various intelligent algorithms have been employed to solve the control parameter optimization of the wind power system with strong coupling and nonlinearity.However,the conventional algorithm has the shortcomings of long optimization time,too large search space,and insufficient global search and local search capabilities,which are impossible to obtain satisfactory control parameters of the RSC of this system.Aiming at the problems existing in conventional intelligent algorithms in optimizing the control parameters of the RSC,this study proposes a quantum parallel multi-layer Monte Carlo optimization algorithm(QPMMCOA)to solve the problem that the conventional intelligent algorithm has too large search space,and insufficient global search and local search capabilities.The QPMMCOA is proposed for accurately finding the optimal solution in a small search space.The QPMMCOA combines qubit probability magnitudes with Monte Carlo random numbers to generate diverse populations and expand the search space.The optimization process of QPMMCOA is divided into rough search,precise search,and re-precise search.These three search processes are completed in the continuously shrinking feasible region,that is,the QPMMCOA seeks the optimal solution by continuously changing and reducing the feasible region.The effectiveness and feasibility of QPMMCOA are confirmed by two benchmark functions.With a wider exploration space and deeper development capabilities,the result of the QPMMCOA is at least 0.51% lower than other algorithms when optimizing the fitness function of the RSC.Compared with other algorithms,the QPMMCOA-based RSC can improve the wind energy conversion efficiency of the wind power system by at least 0.0028%.This study also proposes a deep learning-based quantum parallel multi-layer Monte Carlo optimization algorithm(QPMMCOA-DL)to further solve the problems of long optimization time and insufficient diversity of conventional intelligent algorithms and QPMMCOA in optimizing the control parameters of the RSC.Similar to QPMMCOA,the QPMMCOA-DL mainly introduces mutation operation to improve the population encoding method of the QPMMCOA.The optimization search process of the QPMMCOA-DL is also similar to that of the QPMMCOA,the difference lies in the precise search process,that is,the precise search of QPMMCOA-DL is based on deep learning;this process employs the trained deep neural prediction network in deep learning to replace the fitness function of the RSC;this prediction network can obtain the fitness function value of a population in a very short time,thereby shortening the optimization time in the precise search.The QPMMCOA-DL is verified by two benchmark functions where compares with six intelligent algorithms.The optimization time of the QPMMCOA-DL when searching the parameters of the RSC is 1122.0639 s,which is at most 5.98% of other algorithms.Compared with other algorithms,the QPMMCOA-DL-based RSC can improve the wind energy conversion efficiency of the wind power system by at least 0.0003894%.