Study On LiFePO₄ Battery’s State Of Charge Estimation Algorithm Based On Unscented Kalman Filter

"Safe, energy saving and environmental friendly" is the eternal theme of automotive technology development. Today, as the global energy and environmental issues are becoming severer, new energy vehicles such as electric vehicles are the future trend of the car. As one of the key te chnologies of electric vehicles, the application of battery depends on accurate estimates of battery’s status. State of the battery can be divided into two categories, one can be directly measured, such as voltage, current, temperature, etc.; the other one cannot be directly measured, and can only be estimated by a certain method, which contains the state of charge(SOC) and state of health(SOH). These statuses play an important role during the application of battery.There are many SOC estimation methods, most of which are based on the basic principle of Ampere-hour integration. Although this method is simple, there are two important issues:(1) the initial value of SOC cannot be estimated;(2) current measurement inaccuracies can cause accum ulation of errors. To solve these problems, some methods such as Kalman filtering can be used. As the battery is a highly nonlinear system, using Unscented Kalman Filter(UKF) can get a better result than the Extended Kalman Filter(EKF).In the long-term use process of a battery, aging will occur, which will performance as capacity fading and so on. Therefore, when estimating SOC using the Ampere-hour integration principle, the effect of aging must be considered.In this paper, lithium iron phosphate battery monomer is used for study. At first some various types of equivalent circuit model were compared. Considering the accuracy and complexity of the model, the Thevenin model was selected to describe the dynamic behavior of the battery. The battery model’s parameters were identified by using a HPPC test, and a Beijing Bus Dynamic Stress Test(BBDST) is employed for verification. The results showed that the Thevenin model can describe the dynamic behavior of the battery well with its simple structure and easy method for parameter identification. A constant temperature aging cycle was employed to find out the capacity fading law during the aging process. Then an algorithm based on UKF was realized for battery SOC estimation under Matlab environment. A single working cycle BBDST was employed to verify the algorithm’s effectiveness. By creating initial error and input noise artificially, the robustness and noise immunity of the method can be verified. Simulation results showed that the algorithm can still converge to the true SOC value and follow it even if the initial estimate was inaccurate and the input contained noise. In the late test condition, the estimation error can be controlled to less than 5%. After that, combining a capacity fade model, the algorithm was applied to estimate the battery SOC during different aging level, with a result of stable error less than 5%.