Fault Diagnosis Of Microgrid Based On Optimized Neural Network

With the continuous development of technology and the increase in the penetration rate of distributed power sources,the frequency of failures of the transmission lines and power cables inside the microgrid is also increasing.During the operation of the microgrid,the fault of transmission lines and power cables will cause large-scale power outages and result in very serious economic losses.Therefore,it is necessary to identify and diagnose the faults of the microgrid.This dissertation takes the wind and solar storage microgrid as the research object,and proposes a new fault diagnosis method for the microgrid based on the optimized neural network.The specific research methods are as follows:Firstly,in order to accurately identify the fault types of the microgrid,a fault diagnosis method for microgrid based on extreme learning machine optimized by whale optimization algorithm is proposed.Wavelet packet decomposition is used to extract fault features;then use the WOA to optimize the ELM to establish a diagnostic model to identify and diagnose the fault type.WOA has the characteristics of simple parameter setting,fast learning speed,and strong global optimization ability.The WOA is used to optimize the input weights and hidden layer neuron thresholds of the extreme learning machine,which solves the problem that the random initialization of input weights and hidden layer neuron thresholds easily affects the network performance,and can further improve the learning speed and generalization ability of the network.Conducive to global optimization.The simulation results show that,compared with the traditional neural network,the fault diagnosis model based on the WOA optimized ELM has faster learning speed,stronger generalization ability,and higher recognition accuracy.Secondly,aiming at the problem of insufficient regression ability caused by random selection of input parameters and hidden layer nodes of extreme learning machine,a microgrid fault diagnosis method based on the bayesian optimization algorithm optimized multi-kernel extreme learning machine is proposed.Introduce the kernel function,combine the polynomial kernel function and the Gaussian radial basis kernel function to form a MKELM to establish a fault diagnosis model,and the BOA is used to optimize the relevant parameters of the MKELM.The BOA has strong global optimization capabilities,and can quickly and accurately find the global optimal value through less iterative calculations.The test results show that the proposed method can detect,classify and locate any type of fault in the microgrid with high performance,and its diagnosis accuracy is the highest.Finally,aiming at the problem that the methods proposed in the first two chapters are only applicable to the same operating environment or current microgrid model,a microgrid fault diagnosis method based on the combination of deep transfer learning and long-short-term memory neural network is proposed,which is used to diagnose the faults of microgrid with different structures.First,pre-train the LSTM neural network model based on the source domain data samples and save the relevant parameters;then transfer the parameters in the pre-training model to the domain adaptive network in migration learning to obtain the TL-LSTM model.Marked data(source domain data)and target domain data are used to fine-tune the migration training of the model,and migrate a single microgrid fault diagnosis model to other microgrids of different structures,which improves the generalization ability of the microgrid fault diagnosis model.The test results show that the proposed method can detect and identify any type of faults in microgrids of different structures with high performance,and its diagnosis accuracy is high.Compared with the pre-trained LSTM model before adaptive adjustment,the generalization ability and diagnosis accuracy are improved.