Condition Monitoring And Life Prediction Study Of Mine Lithium-ion Battery

With the development of the coal industry and equipment technology, lithium-ion batteries have been widely applied in monitoring, communication and automation systems of mine electrical equipment. Whether they can provide safe, stable and efficient power input for equipment, directly affects efficient operation of mine and personnel safety. In this paper, large capacity lithium-iron phosphate batteries are chosen as the research object, and a condition monitoring and life prediction system for mine lithium-ion battery is established to enhance stability and security of the power supply system, which realize status monitoring, intelligent management, automatic protection, SOC estimation and RUL prediction, and improve the problems, such as performance decline, low capacity progressively and short life existing in lithium-ion battery.The main research contents include five parts.(1) Analyze the working principle and characteristics of lithium-ion battery, including charge and discharge, rate, cycle characteristics and so on, and summarize influencing factors of SOC and RUL, to determine monitoring parameters.(2) According to standards and function requirements, BMS is designed, including MCU based on STM32, equalization unit, single-bus temperature acquisition unit, high-precision current monitoring and hardware detection and protection unit, and self-inspection of capacity unit.(3) SOC estimation model based on BP neural network is established, considering the intrinsic relationship between SOC and discharge rate, as well as voltage, adopting elastic BP algorithm to improve learning performance. Correctness and effectiveness of this model is proved by measured data based on MATLAB.(4) Considering of the difficulty in building battery model and small sample in application, RVM regression is used for RUL prediction, which uses EM iterative training algorithm to solve the problem that inverse matrix doesn’t exist in algorithms like Mackay iterative, and combines DGM(1,1) with RVM to improve long-term predictability. The effective RUL prediction model is established and proved by training based on measured data and MATLAB afterwards.(5) BMS and upper computer software are developed and implemented. BMS introduces UC/OS-II real time operating system and set different tasks and priorities, such as main program, monitoring and protection task, data transmission and reception task. Upper computer software is based on Lab VIEW, Access and MATLAB, including data management module, alarm module design, SOC estimation and RUL prediction module.