Research On Wind Power Generation System Fault Diagnosis Based On Artificial Immune Algorithm

Wind power system will live up to 20 years, withstanding the test of heat to cold and the extream temperature difference all year around. Due to the law change, variable load and the impact of strong wind gusts, the wind power system will inevitably fail during operation.The technology of wind power system fault diagnosis may detect the faults in advance and reduce the losses caused by the accident. This paper will do an in-depth study on wind power system fault diagnosis technology.Firstly, the principle of doubly-fed wind power generation system has been analyzed.The mathematical model of the doubly-fed wind generator has been established in three different coordinate systems. The model of the doubly-fed wind turbine has been linearly processed by small perturbation analysis method.Then the space vector model of doubly-fed wind turbine has been established.The flux density, the stator parallel branches circulation features, the stator vibration characteristics and the rotor vibration characteristics of the doubly-fed wind turbine during normal operation have all been analyzed.Based on the cause of the malfunction, failure characteristics and failure mechanism of wind power systems analysis, the stator short circuit fault model, the air gap eccentricity fault model and the rotor windings turn short circuit fault model of doubly-fed wind generator have been established.In this paper, the basic principles of biological immune system has been analyzed, the mathematical models in multivariable situations of the immune system have been analyzed under many conditions.Artificial immune algorithm, monoclonal strategy algorithm have been theoretically analyzed and their convergence have been proved.The steps of the immune monoclonal strategy algorithm and polyclonal strategy algorithm have been namely analyzed.This paper describes the fault immune state space of the wind power system and establishes the wind power system fault diagnosis response model based on immune algorithm.Based on this biological immune system theory and operation mechanism, the artificial immune system has been led into wind power system fault intelligent diagnosis. Several improved artificial immune systems that suit wind power system fault diagnoses have been established.This paper has established mathematical model and control strategy of the wind power system network-side converter.It has also established mathematical model and control strategy of the wind power system machine-side converter.The adaptive dynamic clone select algorithm is put forward. The new algorithm is intended to integrate the local searching with the global and the probability evolution searching with the stochastic searching. A fault diagnosis method based on adaptive dynamic clone selection neural network(ADCSNN) is proposed and applied to wind power system converter in this paper. The experimental results show that this method can avoid local minima, has fast convergence and better fault diagnosis performance.Generator is a key component of wind power systems.To the single fault of wind generator, the flux density, parallel branch features of the double-fed wind generator under stator short-circuit fault have been firstly analysed. Based on voltage, current and flux of the doubly-fed wind generator, the mathematical model of a wind generator stator fault has been established. This paper presents a dynamic cloning strategy immune memory algorithm and applies it to only single stator winding fault diagnose of wind generator.Four features are obtained from the wind power generator, and then these features are given to the immune memory dynamic clonal strategy system.The fault related features are extracted as antigen.The immune cloning memory dynamic strategy system has been trained by working status parameter, memory data obtained in the training phase generator sets are used to detect faults.The method uses a memory unit as the category labels,the memory cell is constantly updated by adaptive mean of the population. When the the adaptive degree standard deviation of population is zero, the memory cell does not change to ensure early convergence of the algorithm.Experimental results show that the proposed immune memory dynamically cloning strategy algorithm based on wind power fault diagnosis system has better classification results for the stator windings of a single fault diagnosis.Experimental results also show the proposed method is applicable and effective.When wind power system is running, the stator current fault signal doubly-fed wind generator is relatively weak.The frequency of fault is very close to the base frequency,the magnitude of the fault frequency is also smaller.It is easily submerged by leak fundamental frequency and noise.Under the rotor fault, eccentricity fault, composite fault of rotor and eccentric, the flux density and the stator parallel branches circulation characteristics are theoretically analyzed.On this basis, the wavelet analysis and artificial immune system are combined.To composite fault of wind generator, a new method for compound fault diagnosis of wind power generator eccentricity and rotor based on wavelet analysis-antibody memory clonal algorithm is presented in this paper. The method firstly uses wavelet analysis as a tool to decompose the current signal.The wavelet coefficients by wavelet analysis are calculated, the energy of Doubly-fed wind power generator fault signal by wavelet coefficients are calculated,and then the fault feature after normalized form can be gained.The fault feature is metaphored as antigen.Aantibody is generated by antibody clone memory algorithm.After selection, cloning, mutation, antibody and then reselection, the operations of antibody memory and compression produce new antibodies.These new antibodies may be applied to complex fault diagnosis of wind generator.Then faults can be recognized and classified from fault samples through using the antibody memory clonal algorithm, so as to realize composite fault diagnosis.Experimental results show that the proposed wavelet and memory clonal antibody method have achieved relatively good fault diagnosis capabilities of wind power systemTo the integrated fault of wind power system,this study theoretically analyse the vibration characteristics of wind power system under eccentricity fault, stator fault and rotor windings short circuit fault and composite fault.This paper has proposed a strategy based on adaptive clustering polyclonal method which is applied to integrated fault diagnosis of the wind power generation system.The method defines vibration and current signals of wind power systems as fault feature, it is metaphored as antigen of immune fault diagnosis system.According to the affinity of each antigen and antibody, the adaptation and population size have been adaptively adjusted.The antigens are merged to the determining antibody population.The use of cross operator reorganization, mutation and selection of polyclonal antibodies improve the local search capabilities from the local minima.The integrated fault diagnosis method is applied to the wind power system and has achieved good results.Based on the above studies, this paper also finishes the wind power generator system fault diagnosis software system based on immune algorithm by GE technology and IFIX software, completes system debugging and verification.The fault diagnosis system has been applied to integrated fault diagnosis of wind power system.In this paper, a lot of experiments have validated the higher accuracy and good capabilities of the proposed wind power system fault diagnosis method.Finally, this paper comprehensively summaries the research results, and discusses the further research about wind power system fault diagnosis technology based on artificial immune system.