Design Of Li-ion Battery State Of Charge Estimating Algorithm Based On Unscented Kalman Filter

Electricity is the most important power source nowadays. With the advance of electric storage technology, it is widely used in industry and consumer electronics. Iron Phosphate Li-ion battery technology is a newly developed technology, with the merit of better durable and safety. Nowadays most commercially available batteries are sealed after manufacture, which lack foolproof mechanism against over-charge and overdischarge. To guarantee the optimum performance of battery based energy storage system, battery management is necessary. Estimating the state of charge of Li-ion battery is the basis of battery management.In this thesis, the problem of estimating the state of charge of Iron Phosphate Li-ion Battery was solved. An estimating algorithm which is of accuracy, robustness and adaptivity was designed. The content of this thesis can be divided into the following four aspects.(1). The fundamental theory of battery models were analyzed. Li-ion battery was modeled using a fractional order equivalent circuit model, and the model parameters were identified from battery electric impedance spectrum. Also, the battery was modeled by another typical model-Thevenin model for comparison.(2). The two kind of models were discretized and the respective unscented Kalman filters for estimating state of charge were derived. Also, the algorithm was programmed into MATLAB script.(3). The experiment facility consisting of current and voltage sensing circuits for charge-discharge experiment was built. Power supply and electronic load was used to simulated real load fluctuation of electric vehicles. Voltage and current data was collected and stored on a computer. The reference state of charge was also calculated.(4). The unscented Kalman filter algorithm was verified with the help of MATLAB software. Estimating results were presented in curves. The algorithm was written into the form of conventional programming language and was implemented in embedded system. Also a Li-ion battery state of charge estimating experimental apparatus was also designed.The estimating result indicates that both of the unscented Kalman filter algorithms introduced in this thesis can estimate the state of charge precisely provided that battery models fit the battery well. The algorithms can eliminate initial error on state of charge and the estimated state of charge shall gradually converge to the real figure. When the battery decayed, the capacity shrank, the internal resistance increased, and the battery models did not fit the battery. In these circumstances, the algorithm shall still converge, even though there were errors. When the battery was near fully charged or fully discharged, the estimating error could be within a tolerable range, namely 5%, which fulfill the requirement of avoiding over-charge and over-discharge. The algorithm can be calculate within several millisecond and the requirement of real-time estimation of state of charge can be fulfilled. Embedded system employing these algorithms can also fulfill the requirement of real-time estimation of state of charge. The designed experimental apparatus for state of charge estimation is pragmatic.