在能源和环境问题日益凸显的今天，推进能源绿色“低碳化”和“无碳化”转型将是未来人类社会的长远举措。虚拟电池的出现可以替代真实电池，应用于电驱产品等测试场景。相较于使用真实电池作为电源，虚拟电池无需进行充放电准备，能够模拟真实电池在不同老化程度、温度和荷电状态(State of Charge, SOC)条件下，驱动随机动态变化的负载进行工作，具有测试效率高、测试成本低、快速定位电池状态和安全可靠等优点。基于上述研究背景，本文开展全生命周期虚拟电池研究，从虚拟电池模型构建和仪器开发两个部分，考虑电池老化和温度状态的影响作用，进行相关的研究工作。主要研究内容如下：(1) 对虚拟电池的研究现状和应用场景进行了综述，并综合对比了不同研究之间的差异。分析电池特性的主要影响因素，提出了基于电池全生命周期的虚拟电池研究，并探索了基于数据驱动方法的实现路径。(2) 设计并进行了电池实验，提出了一种基于新型健康因子的电池容量估算方法。该方法从电池充电过程数据中提取表征容量变化的因子，简化预处理步骤，降低了数据采集精度依赖，建立基于支持向量回归(Support Vector Machine, SVR)的容量估算模型。并且，通过对比验证电池老化数据，证明了该模型的可用性。(3) 综合考虑了温度因素和老化因素对电池性能的影响，建立了虚拟电池模型。结合容量估算模型，将电流、电压、温度、SOC作为双向长短期记忆网络(Bi-direction Long Short Term, BiLSTM)的输入，建立了数据驱动的虚拟电池模型。并且设计了优化算法对模型超参数进行全局寻优以达到最佳预测效果。最后在不同老化程度、工况和温度下对虚拟电池模型进行了验证。(4) 进行虚拟电池仪器的功能开发，设计并实现了虚拟电池仪软件。在负载变化情况未知的条件下，虚拟电池仪能够实现和真实电池相同的电学外特性输出。同时，本研究还实现了本地和数字孪生电池平台之间的数据交互。最后，本研究设计了随机工况实验，对虚拟电池与真实电池驱动随机动态变化负载的效果进行了对比验证。
In today‘s world, where energy and environmental issues are increasingly prominent, promoting the green and low-carbon transformation of energy will be a long-term measure for future human society. The emergence of virtual batteries can replace real batteries and be applied in testing scenarios for electric drive products, among others. Compared to using real batteries as a power source, virtual batteries can simulate the end voltage output of real batteries under different aging degrees, temperatures, and SOC(State of Charge) conditions, and drive dynamic loads to work without the need for charge-discharge preparation, thus having the advantages of high testing efficiency, low testing cost, fast location of battery status, and safety and reliability.Based on the above research background, this paper conducts a full life cycle research on virtual batteries, considering the influence of battery aging and temperature states from two parts: virtual battery model construction and instrument development. The main research contents are as follows:(1) A review of the research status and application scenarios of virtual batteries was conducted, and the differences between different studies were comprehensively compared. The main influencing factors of battery characteristics are analyzed, a virtual battery research based on the digital twin battery platform was proposed, and the implementation path based on the data-driven method was explored.(2) Battery experiments were designed and conducted, and a novel capacity estimation method based on health factors was proposed. This method extracts factors that characterize capacity changes from battery charging process data, simplifies preprocessing steps, reduces the dependence on data acquisition accuracy, and establishes a capacity estimation model based on SVR(Support Vector Machine) without special experimental steps. Furthermore, the usability of the model was demonstrated by comparing and validating battery aging data.(3) Considering the influence of temperature and aging on battery performance, a virtual battery model was established. Combining the capacity estimation model, current, voltage, temperature, and SOC were used as the input for BiLSTM(Bi-direction Long Short Term) to establish a data-driven virtual battery model. An optimization algorithm was designed to globally optimize the model hyperparameters to achieve the best prediction effect. Finally, the virtual battery model was verified under different aging degrees, working conditions, and temperatures.(4) The function of the virtual battery instrument was developed, and the virtual battery instrument software was designed and implemented. Under the condition of unknown load changes, the virtual battery instrument can achieve the same output performance as real batteries. At the same time, this research also realized webpage and data interaction between local and cloud platforms. Finally, a random working condition experiment was designed in this study to compare and verify the effects of virtual batteries and real batteries driving random dynamic loads.
