| As an efficient,green and clean transportation tool,electric vehicles are gradually being accepted by people.At present,lithium-ion batteries are often used as the energy supply part of electric vehicles.However,due to the complexity of road traffic,lithium-ion batteries exhibit strong nonlinear degradation characteristics during use,which makes it difficult to accurately estimate their status of health(SOH),which greatly increases Vehicle safety hazards.At present,artificial neural networks are commonly used to estimate the SOH of vehicle-mounted lithium-ion batteries.This paper aims at the problem of BP neural networks that tends to sink into the local minimum when estimating SOH,the simulated annealing(SA)is used to enhance the weights of the BP neural network to further improve the accuracy of the online SOH estimation of lithium-ion batteries.The main work of this paper is as follows:First of all,in order to explore the charging and discharging characteristics of lithium-ion batteries,the internal structure and working mechanism of lithium-ion batteries represented by lithium iron phosphate batteries are analyzed.At the same time,a lithium-ion battery test platform was set up to carry out charging and discharging test experiments on the battery,and to explore the deterioration phenomenon that occurred during the cycle,and to determine the manifestation and cause of the deterioration of the lithium-ion battery.Second,Analyzed the characteristics of the selected lithium iron phosphate battery and constructed its equivalent circuit model.On this basis,the relationship between the battery terminal voltage and the operating current is analyzed,and combined with the battery state change data during the cycle,the health indicator(HI)that is suitable for online measurement and can effectively reflect the degradation state of the lithium iron phosphate battery is extracted,and the voltage drop is constructed.The correlation model between the speed,the drop value of the initial terminal voltage,the number of charge and discharge cycles and the SOH of the lithium iron phosphate battery,and the Pearson correlation analysis method is used to prove their correlation.Third,according to the non-linear degradation characteristics of lithium iron phosphate batteries,combined with the characteristics of artificial neural networks,the BP neural network is selected to test and estimate the battery SOH.Aiming at the characteristics of easily falling into the local minimum in the estimation process,combined with the random disturbance characteristics introduced by the SA algorithm,a BP neural network algorithm based on SA optimization is proposed,and the U-shaped function is used to analyze the effectiveness of the fusion algorithm.Then the extracted HI is used as the input of the estimation model,and MATLAB is used for simulation verification.Finally,the influence of the input HI on the SOH prediction is discussed.Finally,Complete the actual test of the online SOH estimation of on-board lithium-ion batteries in the built online estimation system.The collection of battery status is completed by building an embedded test platform with STM32F407 as the main control chip,and using QT Creator to design a visual interface to realize the human-computer interaction function.In the estimation system,the battery health factor in the online system environment is extracted in real time to realize the online SOH estimation.It also analyzes the possible interference,and proposes several software and hardware filtering methods to improve the accuracy of online SOH estimation.The maximum error obtained at the end is no more than 4%,which proves the portability and accuracy of the fusion algorithm. |
