Online maximum discharge power prediction for lithium-ion batteries with thermal safety constraints

When performing dispatching or regulation through electrical energy storage stations, it is necessary to accurately predict the state of power (SoP) of the lithium-ion batteries. Conventional SoP prediction algorithms almost always use terminal voltage and state of charge as constraints, but temperature has a significant impact on battery SoP, and thus should also be taken into consideration. In this article, online prediction method of battery SoP under temperature constraint is investigated. Firstly, due to the complex nonlinearity among the terminal voltage, temperature, and working current, the SoP cannot be expressed analytically. Instead of using reduced/simplified models to obtain the analytic expression of SoP, as commonly seen in the literature, this work proposes a numeric method that can predict the voltage and temperature through recursion in the model predictive control algorithm. Secondly, considering the possible error accumulation during the recursion, multi-step-ahead Kalman filtering is applied for error correction using the measurable terminal voltage and temperature as criteria. Finally, the effectiveness of the above algorithm is verified with a physical (as opposed to simulated) battery liquid cooling system. The current generated by the above SoP prediction algorithm is applied to an actual 5-string battery pack with a liquid cooling system. After the temperature reaches the preset value for the first time at 707 s in the discharging process, the maximum error between the measured temperature and the set value is 0.2 degrees C, and the average error is -0.02 degrees C. Therefore, the proposed method has great potential to be scaled up to support utility-scale electrical storage applications.