Research Of Remaining Life Intelligent Prediction For Lithium Battery And Application In Building Microgrid

With the gradual penetration of artificial intelligence,the energy crisis is continuously concerned by the international community and the problem of the deterioration of human living environment,the new energy is vigorously developed and used by all countries in the world.Lithium battery is widely used in many fields as a new energy battery because of its many advantages.As the working time of a lithium battery increases,its performance gradually declines,and it will eventually be decommissioned.It is of great significance to predict the remaining life in real time and improve the reliability of the energy system.At the same time,the retired lithium batteries can be recycled step by step,which is in line with the concept of social sustainable development.It is difficult to establish an accurate degradation model for the model-based prediction method,while the data-driven prediction method does not need to establish a degradation model,and only needs to use historical data to predict the remaining life of lithium batteries.Based on the neural network model and datadriven prediction method,the prediction of remaining useful life of lithium battery is studied.The main contents of this paper are as follows:(1)The residual life prediction method of lithium battery based on neural network model is adopted.Aiming at the problem of complex and poor generalization ability of lithium battery degradation model,a neural network model is proposed to predict the remaining life of lithium battery.Because the degradation process of lithium battery runs through the whole life cycle,its capacity data has the characteristics of time series.So designing time series prediction models with Long Short-Term Memory and Gated Recurrent Unit respectively to predict the life of lithium batteries.Through the comparative analysis of experimental results,Gated Recurrent Unit has better prediction effect than Long Short-Term Memory.The software experiment simulation platform is designed to achieve the expected effect of experiment prediction.(2)A prediction method of lithium battery residual life based on Ensemble Empirical Mode Decomposition and Gated Recurrent Unit is proposed.Aiming at the problem that the single model has insufficient accuracy in the prediction of the life of the lithium battery,and at the same time it cannot accurately extract some hidden trends and characteristics within the degradation curve of the lithium battery,so a prediction method using a combined model is proposed.The Ensemble Empirical Mode Decomposition is used to decompose the data set,and then the Gated Recurrent Unit is used to predict subsequence.Finally,the Grid Search optimal super parameter and Adaptive Moment Estimation optimization strategy are used to update the network parameters to predict the life of the lithium battery.The experimental prediction results are compared with other models through the root mean square error and average absolute percentage error.The results verify the feasibility and superiority of this method in the life prediction of lithium battery.The simulation platform is established to realize the intelligent life prediction of lithium battery.(3)Building DC microgrid echelon utilization platform.When the actual capacity of the battery is predicted to decline to 80%?70% of the rated capacity,the battery will be decommissioned.After strict testing and screening,it will be used in the hybrid energy storage system of the DC microgrid of the building to realize the recycling of the lithium battery,which has considerable strategic significance.