| With the large-scale development of renewable energy and the continuous advancement of smart grid construction in various countries around the world,energy storage systems are widely used to solve the intermittent and instability problems of renewable energy.The energy storage system is of great strategic significance for ensuring the security of the power grid,increasing the proportion of renewable energy,improving energy utilization efficiency,and achieving sustainable energy development.The battery management system is one of the core components of the energy storage system,and its effective monitoring and management of the battery pack is directly related to the safety and cost of the entire energy storage system.Due to the large number of single cells in the battery pack,monitoring the performance of single cells is a focus of battery management research.Especially the research on the state of charge(SOC)that reflects the sustainable power supply capacity of the battery and the state of health(SOH)that reflects the degree of battery degradation,both of which are defined quantities and difficult to accurately measure online.Therefore,based on the research on the working principle,failure mechanism,equivalent model and charging and discharging characteristics of the battery,this thesis proposes a battery online monitoring and comprehensive status diagnosis method that is easy to implement in an embedded terminal.Furthermore,in order to overcome the influence of battery type,operating conditions and aging factors,the method based on data-driven model is studied.A joint estimation method of SOC and SOH based on Long Short-Term Memory(LSTM)recurrent neural network with dynamic tracking ability is proposed.Finally,a distributed online monitoring and status diagnosis system for the single cells in the battery pack is designed.The specific work content is as follows:(1)First,research and clarify the limiting conditions of voltage,current,and temperature to ensure the safe operation of the battery pack.Combining the controllable characteristics of the energy storage load and the on-line detection requirements that do not damage the battery,the DC internal resistance detection method is improved,and an on-line detection method for the internal resistance of the battery unit cell under the Thevenin model is proposed.Considering the impact of the performance difference between the single cells on the working performance and safety of the battery pack,it is proposed that after limiting the operating range of battery voltage,current and temperature,the different single cells will be horizontally screened from the three aspects of voltage,temperature and internal resistance.,And conduct a comprehensive health assessment method.(2)Through the gray correlation analysis method,explore the correlation degree of battery SOC,SOH and voltage,current,temperature,impedance parameters and other factors,as well as the coupling between the two.Subsequently,a joint estimation model based on LSTM was established by combining experiments and actual working conditions.This model uses voltage,current,temperature and historical SOH in constant current and constant voltage charging mode as inputs to estimate SOH;uses voltage,current,temperature,average SOH and historical SOC in the full life cycle as inputs to estimate SOC.Using open data sets,from the influence of different discharge rates and temperatures on the model estimation,special working conditions such as intermittent pulse discharge and random discharge,the comparison of the estimation effect of the joint model and the single model,and the comparison of the evaluation parameters between LSTM and BP and Elman networks Experiments have verified the effect of the joint estimation method of SOC and SOH based on LSTM.(3)The distributed network measurement and control system architecture based on CAN bus is adopted to realize the sharing and distributed decision-making of battery state parameter data.With STM32 as the core of the single battery management unit controller,the real-time online monitoring and status diagnosis system of the charge state,health status and fault alarm of each single battery is realized.Using Lab VIEW as the development platform,the design of the host computer for the centralized display and storage of the data information of each distributed battery management unit is realized.In summary,the multi-parameter online detection and comprehensive state assessment method proposed in this thesis does not require complex calculations and is easy to implement in field terminal management equipment.A joint estimation model of SOC and SOH based on LSTM recurrent neural network is proposed,which can more accurately estimate the sustainable power supply capacity of the battery,and provide data for the charging and discharging control strategy of the energy storage battery;it has good generalization ability and can be widely used It is used to evaluate the health status of various types of batteries under different working conditions;using different time scales to meet the estimation requirements while reducing the amount of calculation.This article is based on the CAN bus-based distributed network monitoring system design,which provides a practical solution for the on-line monitoring and status diagnosis of the battery pack and single battery of the microgrid energy storage system. |
