基于神经网络的空间站锂离子蓄电池电性能预测

Electrical Performance Prediction of Space Station Lithium-Ion Batteries Base on Nerual Network

空间站锂离子蓄电池的电性能需要根据任务的变化,做出一定时间的电性能预测。构建了动态神经网络模型,对锂离子蓄电池地面充放电循环数据进行了筛选分类,用于开环和闭环动态神经网络的训练校正和测试,并比较了预测值和实测值,结果显示误差值可控制在1.42%和3.61%之间,此方法可应用于空间站智能电网工程仿真预测和管理实践中。

Predictions should be made for the electrical performance of Lithium-Ion batteries in the space station according to the changes of the tasks.A dynamic neural network model was established and the Lithium-Ion batteries charge and discharge cycles data were selected and categorized for the training, calibration and test of the open-loop and closed-loop dynamic neural network.The predic-ted and measured values were compared.The results showed that the error could be controlled be-tween 1.42%and 3.61%.This method may be used in the smart grid space station engineering simulation, prediction and management.