Study On Remaining Useful Life Prediction Method For Lithium-ion Batteries

In industrial production,because the lithium-ion battery has many features,such as high voltage,small volume,long durability,availability for advantages on many occasions to replace lead-acid batteries,nickel cadmium batteries and so on most other batteries,so rapidly in industry,agriculture,transportation and communication industry and the aerospace field widely the application of.However,the lithium-ion battery if handled improperly will not only cause the battery itself and the components of the damage,or even lead to a series of disaster and engineering accident Therefore,in recent years,scholars at home and abroad are constantly research on lithium-ion battery state of charge and the life estimation,which is more reasonable and effective monitoring of lithium-ion battery running state,avoid all foreseeable disaster accidents,so the residual life of per-lithium-ion battery test will gradually results in the study of hot research scholars at home and abroad the.In view of this,based on the prediction of the remaining life of Li-ion battery,this paper puts forward some methods and tests to achieve the desired results,and makes a further improvement and testing to achieve better results.First of all,lithium-ion battery for industrial application in the actual capacity data usually online prediction is relatively rare,only offline prediction,and obtain some parameters because the online forecasting is not directly obtained from the instrument,can only be estimated by some method,this will inevitably lead to errors.In view of this situation,put forward a kind of extreme learning machine(ELM)network based on indirect methods for predicting the remaining life of the lithium-ion battery for online prediction,analysis of lithium-ion battery remaining life parameters,using the method of indirect prediction,after analyzing,select drop discharge time as lithium-ion battery residual life parameters.The feasibility of using the first partial correlation coefficient method to verify the indirect life characteristic parameters.Finally,the model is established by using the extreme learning machine neural network to predict the remaining life of lithium-ion battery.In addition,the battery output RUL extreme learning machine neural network instability problems based on the further application of genetic ant algorithm on extreme learning machine neural network(Genetic Algorithm Ant Algorithm,GAAA)to optimize the weight and threshold of GAAA,two kinds of optimization algorithm and fusion,it first adopted genetic algorithm(Genetic Algorithms GA)to do the first step of the path optimization,generate the initial pheromone distribution,then using ant colony algorithm(Ant Colony Optimization,ACO)for the further optimization of the path,in order to get better optimization path,the two algorithms are complementary advantages,superior in efficiency of GA,the time efficiency is better than ACO.Finally,the feasibility of using lithium-ion battery data NASA open set to validate the prediction of lithium-ion battery RUL method,and by error analysis of repeated measurements,the results showed that the prediction error is about 5%,that the lithium-ion battery RUL prediction method has good prediction effect,can be very good to achieve the RUL lithium-ion battery industry forecast.