Abstract:

Accurately measuring the voltage of lithium-ion batteries is crucial for improving the reliability of battery state management. Traditional Kalman filter algorithms face limitations, as the Kalman gain does not account for sudden innovation changes, and the state variable estimation process ignores valuable historical data. This leads to unsatisfactory results in the estimation of sampled voltage filtering based on the traditional Kalman algorithm. Therefore, a sampling voltage estimation method based on an improved Kalman filter algorithm incorporating multi-innovation theory has been proposed. This approach extends the single-moment innovation to a multivariate innovation matrix, incorporating information from both current and historical observations. During the system error covariance calculation, the multivariate innovation matrix of current and past observation moments is introduced. For prior estimation corrections, historical state variables, Kalman gains, and the observation moment innovation matrix are added into the process. These improvements refine the traditional algorithm's error covariance and estimation correction procedures, enabling the Kalman gain and state variable estimates to adapt dynamically to changes in innovation. At the same time, a correction factor is introduced to adjust the correction weight of data at different moments. The improved Kalman filter algorithm based on multi-innovation theory is constructed, resolving key limitations of the traditional method by accounting for sudden innovation changes and incorporating historical data. A comparative analysis was performed between the sampled voltage estimated by the improved algorithm and the traditional algorithm. Furthermore, the relationship between voltage and state of charge (SOC) is utilized to calculate the battery SOC. The results show that under constant current discharge conditions, the maximum error of battery sampled voltage decreases from 8.09 mV to 3.71 mV. Moreover, the SOC error calculated using the sampled voltage is significantly reduced. The improved method reduces the average error and root mean square error (RMSE) for sampled voltage and SOC. These results validate the effectiveness of the improved Kalman filter algorithm and provide new ideas for improving the accuracy of battery sampled voltage filtering estimation and SOC calculation.

Key words: Li-ion battery, sampling voltage, multi-innovation theory, Kalman filter algorithm, state of charge(SOC)